Proton/jxrlib/image/decode/strdec.c
2020-09-29 14:29:06 -05:00

3631 lines
140 KiB
C

//*@@@+++@@@@******************************************************************
//
// Copyright © Microsoft Corp.
// All rights reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are met:
//
// • Redistributions of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
// • Redistributions in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other materials provided with the distribution.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
// ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
// LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
// CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
// SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
// INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
// CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
// ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
// POSSIBILITY OF SUCH DAMAGE.
//
//*@@@---@@@@******************************************************************
#include "strcodec.h"
#include "decode.h"
#include "strTransform.h"
#include <math.h>
#include "perfTimer.h"
#ifdef MEM_TRACE
#define TRACE_MALLOC 1
#define TRACE_NEW 0
#define TRACE_HEAP 0
#include "memtrace.h"
#endif
#ifdef X86OPT_INLINE
#define _FORCEINLINE __forceinline
#else // X86OPT_INLINE
#define _FORCEINLINE
#endif // X86OPT_INLINE
#if defined(WMP_OPT_SSE2) || defined(WMP_OPT_CC_DEC) || defined(WMP_OPT_TRFM_DEC)
void StrDecOpt(CWMImageStrCodec* pSC);
#endif // OPT defined
Int processMacroblockDec(CWMImageStrCodec *);
U8 readQuantizerSB(U8 pQPIndex[MAX_CHANNELS], SimpleBitIO * pIO, size_t cChannel)
{
U8 cChMode = 0;
if(cChannel >= MAX_CHANNELS)
return 0;
if(cChannel > 1)
cChMode = (U8)getBit32_SB(pIO, 2); // Channel mode
pQPIndex[0] = (U8)getBit32_SB(pIO, 8); // Y
if(cChMode == 1) // MIXED
pQPIndex[1] = (U8)getBit32_SB(pIO, 8); // UV
else if(cChMode > 0){ // INDEPENDENT
size_t i;
for(i = 1; i < cChannel; i ++)
#pragma prefast(suppress: __WARNING_UNRELATED_LOOP_TERMINATION_NO_SIZEEXPR, "PREfast false alarm: 1 <= i < MAX_CHANNELS, no buffer over/underrun!")
pQPIndex[i] = (U8)getBit32_SB(pIO, 8); // UV
}
return cChMode;
}
U8 readQuantizer(CWMIQuantizer * pQuantizer[MAX_CHANNELS], BitIOInfo * pIO, size_t cChannel, size_t iPos)
{
U8 cChMode = 0;
if(cChannel > 1)
cChMode = (U8)getBit16(pIO, 2); // Channel mode
pQuantizer[0][iPos].iIndex = (U8)getBit16(pIO, 8); // Y
if(cChMode == 1) // MIXED
pQuantizer[1][iPos].iIndex = (U8)getBit16(pIO, 8); // UV
else if(cChMode > 0){ // INDEPENDENT
size_t i;
for(i = 1; i < cChannel; i ++)
pQuantizer[i][iPos].iIndex = (U8)getBit16(pIO, 8); // UV
}
return cChMode;
}
// packet header: 00000000 00000000 00000001 ?????xxx
// xxx: 000(spatial) 001(DC) 010(AD) 011(AC) 100(FL) 101-111(reserved)
// ?????: (iTileY * cNumOfSliceV + iTileX) % 32
Int readPacketHeader(BitIOInfo * pIO, U8 ptPacketType, U8 pID)
{
UNREFERENCED_PARAMETER( ptPacketType );
UNREFERENCED_PARAMETER( pID );
if(getBit16(pIO, 8) != 0 || getBit16(pIO, 8) != 0 || getBit16(pIO, 8) != 1)
return ICERR_ERROR;
getBit16(pIO, 8);
return ICERR_OK;
}
Int readTileHeaderDC(CWMImageStrCodec * pSC, BitIOInfo * pIO)
{
if((pSC->m_param.uQPMode & 1) != 0){ // not DC uniform
size_t iTile;
CWMITile * pTile = pSC->pTile + pSC->cTileColumn;
if(pSC->cTileRow + pSC->cTileColumn == 0) // allocate DC QP info
for(iTile = 0; iTile <= pSC->WMISCP.cNumOfSliceMinus1V; iTile ++)
if(allocateQuantizer(pSC->pTile[iTile].pQuantizerDC, pSC->m_param.cNumChannels, 1) != ICERR_OK)
return ICERR_ERROR;
pTile->cChModeDC = readQuantizer(pTile->pQuantizerDC, pIO, pSC->m_param.cNumChannels, 0);
formatQuantizer(pTile->pQuantizerDC, pTile->cChModeDC, pSC->m_param.cNumChannels, 0, TRUE, pSC->m_param.bScaledArith);
}
return ICERR_OK;
}
Int readTileHeaderLP(CWMImageStrCodec * pSC, BitIOInfo * pIO)
{
if(pSC->WMISCP.sbSubband != SB_DC_ONLY && (pSC->m_param.uQPMode & 2) != 0){ // not LP uniform
CWMITile * pTile = pSC->pTile + pSC->cTileColumn;
U8 i;
pTile->bUseDC = (getBit16(pIO, 1) == 1 ? TRUE : FALSE);
pTile->cBitsLP = 0;
pTile->cNumQPLP = 1;
if(pSC->cTileRow > 0)
freeQuantizer(pTile->pQuantizerLP);
if(pTile->bUseDC == TRUE){
if(allocateQuantizer(pTile->pQuantizerLP, pSC->m_param.cNumChannels, pTile->cNumQPLP) != ICERR_OK)
return ICERR_ERROR;
useDCQuantizer(pSC, pSC->cTileColumn);
}
else{
pTile->cNumQPLP = (U8)getBit16(pIO, 4) + 1;
pTile->cBitsLP = dquantBits(pTile->cNumQPLP);
if(allocateQuantizer(pTile->pQuantizerLP, pSC->m_param.cNumChannels, pTile->cNumQPLP) != ICERR_OK)
return ICERR_ERROR;
for(i = 0; i < pTile->cNumQPLP; i ++){
pTile->cChModeLP[i] = readQuantizer(pTile->pQuantizerLP, pIO, pSC->m_param.cNumChannels, i);
formatQuantizer(pTile->pQuantizerLP, pTile->cChModeLP[i], pSC->m_param.cNumChannels, i, TRUE, pSC->m_param.bScaledArith);
}
}
}
return ICERR_OK;
}
Int readTileHeaderHP(CWMImageStrCodec * pSC, BitIOInfo * pIO)
{
if(pSC->WMISCP.sbSubband != SB_DC_ONLY && pSC->WMISCP.sbSubband != SB_NO_HIGHPASS && (pSC->m_param.uQPMode & 4) != 0){ // not HP uniform
CWMITile * pTile = pSC->pTile + pSC->cTileColumn;
U8 i;
pTile->bUseLP = (getBit16(pIO, 1) == 1 ? TRUE : FALSE);
pTile->cBitsHP = 0;
pTile->cNumQPHP = 1;
if(pSC->cTileRow > 0)
freeQuantizer(pTile->pQuantizerHP);
if(pTile->bUseLP == TRUE){
pTile->cNumQPHP = pTile->cNumQPLP;
if(allocateQuantizer(pTile->pQuantizerHP, pSC->m_param.cNumChannels, pTile->cNumQPHP) != ICERR_OK)
return ICERR_ERROR;
useLPQuantizer(pSC, pTile->cNumQPHP, pSC->cTileColumn);
}
else{
pTile->cNumQPHP = (U8)getBit16(pIO, 4) + 1;
pTile->cBitsHP = dquantBits(pTile->cNumQPHP);
if(allocateQuantizer(pTile->pQuantizerHP, pSC->m_param.cNumChannels, pTile->cNumQPHP) != ICERR_OK)
return ICERR_ERROR;
for(i = 0; i < pTile->cNumQPHP; i ++){
pTile->cChModeHP[i] = readQuantizer(pTile->pQuantizerHP, pIO, pSC->m_param.cNumChannels, i);
formatQuantizer(pTile->pQuantizerHP, pTile->cChModeHP[i], pSC->m_param.cNumChannels, i, FALSE, pSC->m_param.bScaledArith);
}
}
}
return ICERR_OK;
}
Int readPackets(CWMImageStrCodec * pSC)
{
if(pSC->cColumn == 0 && pSC->cRow == pSC->WMISCP.uiTileY[pSC->cTileRow]){ // start of a new horizontal slice
size_t k;
if (pSC->m_bSecondary) {
if(pSC->cNumBitIO > 0){
for(k = 0; k <= pSC->WMISCP.cNumOfSliceMinus1V; k ++){
// reset coding contexts
ResetCodingContextDec(&pSC->m_pCodingContext[k]);
}
}
else{ // for multiple decoding calls!
ResetCodingContextDec(&pSC->m_pCodingContext[0]);
}
}
else {
// get sizes of each packet and update index table
for(k = 0; k < pSC->cNumBitIO; k ++){
if(pSC->ppWStream != NULL){ // new API
unsigned cBands = (pSC->WMISCP.bfBitstreamFormat == SPATIAL ? 1 : pSC->cSB);
struct WMPStream ** ppWS = pSC->ppWStream + (pSC->WMISCP.cNumOfSliceMinus1V + 1) * pSC->cTileRow * cBands
+ k / cBands * cBands + (k % cBands);
if(pSC->cTileRow > 0 && pSC->m_ppBitIO[k]->pWS != NULL) // attached to the same packet of the tile on top
detachISRead(pSC, pSC->m_ppBitIO[k]); // detach it
if(ppWS[0] != NULL)
attachISRead(pSC->m_ppBitIO[k], ppWS[0], pSC); // need to attach it
}
else{
if(pSC->cTileRow > 0)
detachISRead(pSC, pSC->m_ppBitIO[k]);
pSC->WMISCP.pWStream->SetPos(pSC->WMISCP.pWStream, pSC->pIndexTable[pSC->cNumBitIO * pSC->cTileRow + k] + pSC->cHeaderSize);
attachISRead(pSC->m_ppBitIO[k], pSC->WMISCP.pWStream, pSC);
}
}
if(pSC->cNumBitIO == 0){
detachISRead(pSC, pSC->pIOHeader);
if(pSC->ppWStream != NULL){// new API
attachISRead(pSC->pIOHeader, pSC->ppWStream[0], pSC); // need to attach it
}
else{
pSC->WMISCP.pWStream->SetPos(pSC->WMISCP.pWStream, pSC->cHeaderSize);
attachISRead(pSC->pIOHeader, pSC->WMISCP.pWStream, pSC);
}
}
for(k = 0; k <= pSC->WMISCP.cNumOfSliceMinus1V; k ++){
U8 pID = (U8)((pSC->cTileRow * (pSC->WMISCP.cNumOfSliceMinus1V + 1) + k) & 0x1F);
// read packet header
if(pSC->WMISCP.bfBitstreamFormat == SPATIAL){
BitIOInfo * pIO = (pSC->cNumBitIO == 0 ? pSC->pIOHeader : pSC->m_ppBitIO[k]);
if(pIO->pWS == NULL || readPacketHeader(pIO, 0, pID) != ICERR_OK)
return ICERR_ERROR;
pSC->m_pCodingContext[k].m_iTrimFlexBits = (pSC->m_param.bTrimFlexbitsFlag) ? getBit16(pIO, 4) : 0;
}
else{
if(pSC->m_ppBitIO[k * pSC->cSB + 0] == NULL || readPacketHeader(pSC->m_ppBitIO[k * pSC->cSB + 0], 1, pID) != ICERR_OK)
return ICERR_ERROR;
if(pSC->cSB > 1){
if(pSC->m_ppBitIO[k * pSC->cSB + 1] == NULL || readPacketHeader(pSC->m_ppBitIO[k * pSC->cSB + 1], 2, pID) != ICERR_OK)
return ICERR_ERROR;
}
if(pSC->cSB > 2){
if(pSC->m_ppBitIO[k * pSC->cSB + 2] == NULL || readPacketHeader(pSC->m_ppBitIO[k * pSC->cSB + 2], 3, pID) != ICERR_OK)
return ICERR_ERROR;
// readTileHeaderHP(pSC, pSC->m_ppBitIO[k * pSC->cSB + 2]);
}
if(pSC->cSB > 3){
if(pSC->m_ppBitIO[k * pSC->cSB + 3] == NULL)
return ICERR_ERROR;
readPacketHeader(pSC->m_ppBitIO[k * pSC->cSB + 3], 4, pID); // bad flexbits packet doesn't generate an error
pSC->m_pCodingContext[k].m_iTrimFlexBits = (pSC->m_param.bTrimFlexbitsFlag) ? getBit16(pSC->m_ppBitIO[k * pSC->cSB + 3], 4) : 0;
}
}
// reset coding contexts
ResetCodingContextDec(&pSC->m_pCodingContext[k]);
}
}
}
if(pSC->m_bCtxLeft && pSC->m_bCtxTop && pSC->m_bSecondary == FALSE){
CCodingContext *pContext = &pSC->m_pCodingContext[pSC->cTileColumn];
readTileHeaderDC(pSC, pContext->m_pIODC);
if(pSC->m_pNextSC != NULL)
readTileHeaderDC(pSC->m_pNextSC, pContext->m_pIODC);
if(pSC->cSB > 1){
readTileHeaderLP(pSC, pContext->m_pIOLP);
if(pSC->m_pNextSC != NULL)
readTileHeaderLP(pSC->m_pNextSC, pContext->m_pIOLP);
}
if(pSC->cSB > 2){
readTileHeaderHP(pSC, pContext->m_pIOAC);
if(pSC->m_pNextSC != NULL)
readTileHeaderHP(pSC->m_pNextSC, pContext->m_pIOAC);
}
}
return ICERR_OK;
}
/* inverse transform and overlap possible part of a macroblock */
Int processMacroblockDec(CWMImageStrCodec * pSC)
{
const OVERLAP olOverlap = pSC->WMISCP.olOverlap;
// const Bool left = (pSC->cColumn == 0);
const Bool /*top = (pSC->cRow == 0),*/ bottom = (pSC->cRow == pSC->cmbHeight);
const Bool bottomORright = (bottom || pSC->cColumn == pSC->cmbWidth);
// const size_t mbWidth = pSC->cmbWidth, mbX = pSC->cColumn;
// Int iQIndex = 0;
ERR_CODE result = ICERR_OK;
size_t j, jend = (pSC->m_pNextSC != NULL);
for (j = 0; j <= jend; j++) {
if(!bottomORright){
CCodingContext *pContext;
getTilePos(pSC, pSC->cColumn, pSC->cRow);
if(jend){
pSC->m_pNextSC->cTileColumn = pSC->cTileColumn;
pSC->m_pNextSC->cTileRow = pSC->cTileRow;
}
pContext = &pSC->m_pCodingContext[pSC->cTileColumn];
if(readPackets(pSC) != ICERR_OK)
return ICERR_ERROR;
// check if we need to do entropy decode
if(!pSC->m_Dparam->bDecodeFullFrame){
if(pSC->cColumn == pSC->WMISCP.uiTileX[pSC->cTileColumn]){ // switching to a new tile
size_t rLeft = pSC->m_Dparam->cROILeftX, rRight = pSC->m_Dparam->cROIRightX;
size_t rTop = pSC->m_Dparam->cROITopY, rBottom = pSC->m_Dparam->cROIBottomY;
size_t rExt = (olOverlap == OL_NONE ? 0 : olOverlap == OL_ONE ? 2 : 10);
size_t tLeft = pSC->cColumn * 16, tTop = pSC->WMISCP.uiTileY[pSC->cTileRow] * 16;
size_t tRight = (pSC->cTileColumn != pSC->WMISCP.cNumOfSliceMinus1V ? pSC->WMISCP.uiTileX[pSC->cTileColumn + 1] : pSC->cmbWidth) * 16;
size_t tBottom = (pSC->cTileRow != pSC->WMISCP.cNumOfSliceMinus1H ? pSC->WMISCP.uiTileY[pSC->cTileRow + 1] : pSC->cmbHeight) * 16;
// tile overlaps with ROI?
pContext->m_bInROI = ((rLeft >= tRight + rExt || rTop >= tBottom + rExt || tLeft > rRight + rExt ||
tTop > rBottom + rExt || pSC->cRow * 16 > rBottom + rExt) ? FALSE : TRUE);
}
}
if(pSC->m_Dparam->bDecodeFullFrame || pContext->m_bInROI){
if ((result = DecodeMacroblockDC(pSC, pContext, (Int)pSC->cColumn, (Int)pSC->cRow)) != ICERR_OK)
return result;
if(pSC->m_Dparam->bDecodeLP){
if ((result = DecodeMacroblockLowpass(pSC, pContext, (Int)pSC->cColumn, (Int)pSC->cRow)) != ICERR_OK)
return result;
}
predDCACDec(pSC);
dequantizeMacroblock(pSC);
if(pSC->m_Dparam->bDecodeHP){
if ((result = DecodeMacroblockHighpass(pSC, pContext, (Int)pSC->cColumn, (Int)pSC->cRow)) != ICERR_OK)
return result;
predACDec(pSC);
}
/* keep necessary info for future prediction */
updatePredInfo(pSC, &pSC->MBInfo, (Int)pSC->cColumn, pSC->m_param.cfColorFormat);
}
}
if((!pSC->m_Dparam->bDecodeFullFrame) &&
((pSC->cColumn * 16 > pSC->m_Dparam->cROIRightX + 25) || (pSC->cColumn * 16 + 25 < pSC->m_Dparam->cROILeftX)
|| (pSC->cRow * 16 > pSC->m_Dparam->cROIBottomY + 25) || (pSC->cRow * 16 + 25 < pSC->m_Dparam->cROITopY)))
{
// do nothing
}
else {
pSC->Transform(pSC);
}
if (jend) {
pSC->m_pNextSC->cRow = pSC->cRow;
pSC->m_pNextSC->cColumn = pSC->cColumn;
pSC = pSC->m_pNextSC;
}
}
return result;
}
//================================================================
// Inverse Color Conversion
//#define _ICC1(r, g, b) (g^=b^=g^=b, r^=g^=r^=g, b += ((g) >> 1), r += ((g) >> 1), g -= (b+3*r+2) >> 2)
//#define _ICC(r, g, b) (g^=b^=g^=b, r^=g^=r^=g, b += ((g) >> 1), r += ((g) >> 1), g -= (b+3*r+2) >> 2)
//================================================================
//#define _ICC1(r, g, b) r -= (g >> 1), g += r, r -= ((b + 1) >> 1), b += r
//#define _ICC(r, g, b) r -= (g >> 1), g += r, r -= (b >> 1), b += r
#define _ICC(r, g, b) (g -= ((r + 0) >> 1), r -= ((b + 1) >> 1) - g, b += r)
#define _ICC_CMYK(c, m, y, k) (k -= ((m + 1) >> 1), m -= (c >> 1) - k, c -= ((y + 1) >> 1) - m, y += c)
#define _CLIP2(l, v, h) ((v) < (l) ? (l) : ((h) < (v) ? (h) : (v)))
#define _CLIP8(v) ((U8)_CLIP2(0, v, 255))
#define _CLIP16(v) ((I16)_CLIP2(-32768, v, 32767))
#define _CLIPU16(v) ((U16)_CLIP2(0, v, 65535))
#define min(a,b) (((a) < (b)) ? (a) : (b))
//inverseConvert: Inverse conversion from float RGB to RGBE
static _FORCEINLINE void inverseConvert (PixelI iF, U8 *pRGB, U8 *pE)
{
if (iF <= 0) {
*pRGB = *pE = 0;
}
else if ((iF >> 7) > 1) {
/** normal form **/
*pE = (U8) (iF >> 7); //+ 1;
*pRGB = (iF & 0x7f) | 0x80;
}
else {
/** denormal form **/
*pE = 1;
*pRGB = (U8) iF;
}
}
#ifdef __ANSI__
#define max(a,b) ((a) > (b) ? (a) : (b))
#endif // __ANSI__
static _FORCEINLINE void inverseConvertRGBE (PixelI iFr, PixelI iFg, PixelI iFb, U8 *pR, U8 *pG, U8 *pB, U8 *pE)
{
U8 iShift;
U8 pR_E, pG_E, pB_E;
inverseConvert (iFr, pR, &pR_E);
inverseConvert (iFg, pG, &pG_E);
inverseConvert (iFb, pB, &pB_E);
*pE = max(max(pR_E, pG_E), pB_E);
if(*pE > pR_E){
iShift = (*pE - pR_E);
*pR = (U8)((((int)*pR) * 2 + 1) >> (iShift + 1));
}
if(*pE > pG_E){
iShift = (*pE - pG_E);
*pG = (U8)((((int)*pG) * 2 + 1) >> (iShift + 1));
}
if(*pE > pB_E){
iShift = (*pE - pB_E);
*pB = (U8)((((int)*pB) * 2 + 1) >> (iShift + 1));
}
}
//pixel to float 32!
static _FORCEINLINE float pixel2float(PixelI _h, const char _c, const unsigned char _lm)
{
union uif
{
I32 i;
float f;
} x;
I32 s, iTempH, m, e, lmshift = (1 << _lm);
// assert (_c <= 127);
iTempH = (I32) _h ;
s = (iTempH >> 31);
iTempH = (iTempH ^ s) - s; // abs(iTempH)
e = (U32) iTempH >> _lm;// & ((1 << (31 - _lm)) - 1);
m = (iTempH & (lmshift - 1)) | lmshift; // actual mantissa, with normalizer
if (e == 0) { // denormal land
m ^= lmshift; // normalizer removed
e = 1; // actual exponent
}
e += (127 - _c);
while (m < lmshift && e > 1 && m > 0) { // denormal originally, see if normal is possible
e--;
m <<= 1;
}
if (m < lmshift) // truly denormal
e = 0;
else
m ^= lmshift;
m <<= (23 - _lm);
x.i = (s & 0x80000000) | (e << 23) | m;
return x.f;
}
//convert Half-16 to internal format, only need to handle sign bit
static _FORCEINLINE U16 backwardHalf (PixelI hHalf)
{
PixelI s;
s = hHalf >> 31;
hHalf = ((hHalf & 0x7fff) ^ s) - s; // don't worry about overflow
return (U16) hHalf;
}
Void interpolateUV(CWMImageStrCodec * pSC)
{
const COLORFORMAT cfExt = pSC->WMII.cfColorFormat;
const size_t cWidth = pSC->cmbWidth * 16;
PixelI * pSrcU = pSC->a0MBbuffer[1], * pSrcV = pSC->a0MBbuffer[2];
PixelI * pDstU = pSC->pResU, * pDstV = pSC->pResV;
size_t iRow, iColumn;
size_t iIdxS = 0, iIdxD = 0;
if(pSC->m_param.cfColorFormat == YUV_422){ // 422 => 444, interpolate horizontally
for(iRow = 0; iRow < 16; iRow ++){
for(iColumn = 0; iColumn < cWidth; iColumn += 2){
iIdxS = ((iColumn >> 4) << 7) + idxCC[iRow][(iColumn >> 1) & 7];
iIdxD = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
// copy over
pDstU[iIdxD] = pSrcU[iIdxS];
pDstV[iIdxD] = pSrcV[iIdxS];
if(iColumn > 0){
size_t iL = iColumn - 2, iIdxL = ((iL >> 4) << 8) + idxCC[iRow][iL & 15];
size_t iC = iColumn - 1, iIdxC = ((iC >> 4) << 8) + idxCC[iRow][iC & 15];
// interpolate
pDstU[iIdxC] = ((pDstU[iIdxL] + pDstU[iIdxD] + 1) >> 1);
pDstV[iIdxC] = ((pDstV[iIdxL] + pDstV[iIdxD] + 1) >> 1);
}
}
//last pixel
iIdxS = (((iColumn - 1) >> 4) << 8) + idxCC[iRow][(iColumn - 1) & 15];
pDstU[iIdxS] = pDstU[iIdxD];
pDstV[iIdxS] = pDstV[iIdxD];
}
}
else{ // 420 => 422 or 444, interpolate vertically
const size_t cShift = (cfExt == YUV_422 ? 3 : 4);
for(iColumn = 0; iColumn < cWidth; iColumn += 2){
const size_t cMB = ((iColumn >> 4) << (4 + cShift)), cPix = (iColumn >> (4 - cShift)) & ((1 << cShift) - 1);
for(iRow = 0; iRow < 16; iRow += 2){
iIdxS = ((iColumn >> 4) << 6) + idxCC_420[iRow >> 1][(iColumn >> 1) & 7];
iIdxD = cMB + idxCC[iRow][cPix];
// copy over
pDstU[iIdxD] = pSrcU[iIdxS];
pDstV[iIdxD] = pSrcV[iIdxS];
if(iRow > 0){
size_t iIdxT = cMB + idxCC[iRow - 2][cPix];
size_t iIdxC = cMB + idxCC[iRow - 1][cPix];
// interpolate
pDstU[iIdxC] = ((pDstU[iIdxT] + pDstU[iIdxD] + 1) >> 1);
pDstV[iIdxC] = ((pDstV[iIdxT] + pDstV[iIdxD] + 1) >> 1);
}
}
//last row
iIdxS = cMB + idxCC[15][cPix];
if(pSC->cRow == pSC->cmbHeight){ // image boundary
pDstU[iIdxS] = pDstU[iIdxD];
pDstV[iIdxS] = pDstV[iIdxD];
}
else{ // need next MB row
size_t iIdxB = ((iColumn >> 4) << 6) + idxCC_420[0][(iColumn >> 1) & 7];
pDstU[iIdxS] = ((pSC->a1MBbuffer[1][iIdxB] + pDstU[iIdxD] + 1) >> 1);
pDstV[iIdxS] = ((pSC->a1MBbuffer[2][iIdxB] + pDstV[iIdxD] + 1) >> 1);
}
}
if(cfExt != YUV_422){ // 420 => 444, interpolate horizontally
for(iRow = 0; iRow < 16; iRow ++){
for(iColumn = 1; iColumn < cWidth - 2; iColumn += 2){
size_t iIdxL = (((iColumn - 1) >> 4) << 8) + idxCC[iRow][(iColumn - 1) & 15];
iIdxD = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
iIdxS = (((iColumn + 1) >> 4) << 8) + idxCC[iRow][(iColumn + 1) & 15];
pDstU[iIdxD] = ((pDstU[iIdxS] + pDstU[iIdxL] + 1) >> 1);
pDstV[iIdxD] = ((pDstV[iIdxS] + pDstV[iIdxL] + 1) >> 1);
}
// last pixel
iIdxD = (((cWidth - 1) >> 4) << 8) + idxCC[iRow][(cWidth - 1) & 15];
pDstU[iIdxD] = pDstU[iIdxS];
pDstV[iIdxD] = pDstV[iIdxS];
}
}
}
}
// write one MB row of Y_ONLY/CF_ALPHA/YUV_444/N_CHANNEL to output buffer
Void outputNChannel(CWMImageStrCodec * pSC, size_t iFirstRow, size_t iFirstColumn, size_t cWidth, size_t cHeight, size_t iShift, PixelI iBias)
{
const CWMImageInfo* pII = &pSC->WMII;
const size_t cChannel = pII->cfColorFormat == Y_ONLY ? 1 : pSC->WMISCP.cChannel;
// const U8 cbChannels[BDB_MAX] = {-1, 1, 2, 2, 2, 4, 4, -1, -1, };
const U8 nLen = pSC->WMISCP.nLenMantissaOrShift;
const I8 nExpBias = pSC->WMISCP.nExpBias;
PixelI * pChannel[16];
size_t iChannel, iRow, iColumn;
size_t * pOffsetX = pSC->m_Dparam->pOffsetX, * pOffsetY = pSC->m_Dparam->pOffsetY + (pSC->cRow - 1) * 16, iY;
assert(cChannel <= 16);
for(iChannel = 0; iChannel < cChannel; iChannel ++)
pChannel[iChannel & 15] = pSC->a0MBbuffer[iChannel];
if(pSC->m_bUVResolutionChange)
pChannel[1] = pSC->pResU, pChannel[2] = pSC->pResV;
switch(pSC->WMII.bdBitDepth){
case BD_8:
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
U8 * pDst = (U8 *)pSC->WMIBI.pv + iY + pOffsetX[iColumn];
for(iChannel = 0; iChannel < cChannel; iChannel ++){
PixelI p = ((pChannel[iChannel & 15][((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iBias) >> iShift);
pDst[iChannel] = _CLIP8(p);
}
}
}
break;
case BD_16:
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
U16 * pDst = (U16 *)pSC->WMIBI.pv + iY + pOffsetX[iColumn];
for(iChannel = 0; iChannel < cChannel; iChannel ++){
PixelI p = ((pChannel[iChannel & 15][((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iBias) >> iShift);
p <<= nLen;
pDst[iChannel] = _CLIPU16(p);
}
}
}
break;
case BD_16S:
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
I16 * pDst = (I16 *)pSC->WMIBI.pv + iY + pOffsetX[iColumn];
for(iChannel = 0; iChannel < cChannel; iChannel ++){
PixelI p = ((pChannel[iChannel & 15][((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iBias) >> iShift);
p <<= nLen;
pDst[iChannel] = _CLIP16(p);
}
}
}
break;
case BD_16F:
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
U16 * pDst = (U16 *)pSC->WMIBI.pv + iY + pOffsetX[iColumn];
for(iChannel = 0; iChannel < cChannel; iChannel ++){
PixelI p = ((pChannel[iChannel & 15][((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf]] + iBias) >> iShift);
pDst[iChannel] = backwardHalf(p);
}
}
}
break;
case BD_32:
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
U32 * pDst = (U32 *)pSC->WMIBI.pv + iY + pOffsetX[iColumn];
for(iChannel = 0; iChannel < cChannel; iChannel ++){
PixelI p = ((pChannel[iChannel & 15][((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf]] + iBias) >> iShift);
p <<= nLen;
pDst[iChannel] = (U32)(p);
}
}
}
break;
case BD_32S:
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
I32 * pDst = (I32 *)pSC->WMIBI.pv + iY + pOffsetX[iColumn];
for(iChannel = 0; iChannel < cChannel; iChannel ++){
PixelI p = ((pChannel[iChannel & 15][((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf]] + iBias) >> iShift);
p <<= nLen;
pDst[iChannel] = (I32)(p);
}
}
}
break;
case BD_32F:
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
float * pDst = (float *)pSC->WMIBI.pv + iY + pOffsetX[iColumn];
for(iChannel = 0; iChannel < cChannel; iChannel ++){
PixelI p = ((pChannel[iChannel & 15][((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf]] + iBias) >> iShift);
pDst[iChannel] = pixel2float (p, nExpBias, nLen);
}
}
}
break;
default:
assert(0);
break;
}
}
static void fixup_Y_ONLY_to_Others(
const CWMImageStrCodec* pSC,
const CWMImageBufferInfo* pBI)
{
const CWMImageInfo* pII = &pSC->WMII;
const CWMIStrCodecParam* pSCP = &pSC->WMISCP;
size_t cWidth = 0, cHeight = 0;
size_t idxY = 0, idxX = 0;
if (CF_RGB != pII->cfColorFormat || Y_ONLY != pSCP->cfColorFormat)
return;
cWidth = 0 != pII->cROIWidth ? pII->cROIWidth : pII->cWidth;
cHeight = 0 != pII->cROIHeight ? pII->cROIHeight : pII->cHeight;
#define fixup(type, nCh) \
for (idxY = 0; idxY < cHeight; ++idxY) \
{ \
type * pT = (type *)((U8*)pBI->pv + pBI->cbStride * idxY); \
for (idxX = 0; idxX < cWidth; ++idxX) \
{ \
pT[2] = pT[1] = pT[0]; \
pT += nCh; \
} \
} \
break
switch (pII->bdBitDepth)
{
case BD_8:
fixup(U8, (pII->cBitsPerUnit >> 3));
break;
case BD_16:
case BD_16S:
case BD_16F:
fixup(U16, (pII->cBitsPerUnit >> 3) / sizeof(U16));
break;
case BD_32:
case BD_32S:
case BD_32F:
fixup(U32, (pII->cBitsPerUnit >> 3) / sizeof(float));
break;
case BD_5:
case BD_10:
case BD_565:
default:
break;
}
}
// centralized alpha channel color conversion, small perf penalty
Int outputMBRowAlpha(CWMImageStrCodec * pSC)
{
if(pSC->WMII.bdBitDepth == BD_8 && pSC->WMISCP.cfColorFormat == CF_RGB) // has been taken care of and optimized out
return ICERR_OK;
if(pSC->m_bSecondary == FALSE && pSC->m_pNextSC != NULL){ // with alpha channel
const BITDEPTH_BITS bd = pSC->WMII.bdBitDepth;
const PixelI iShift = (pSC->m_param.bScaledArith ? SHIFTZERO + QPFRACBITS : 0);
const size_t cHeight = min((pSC->m_Dparam->cROIBottomY + 1) - (pSC->cRow - 1) * 16, 16);
const size_t cWidth = (pSC->m_Dparam->cROIRightX + 1);
const size_t iFirstRow = ((pSC->cRow - 1) * 16 > pSC->m_Dparam->cROITopY ? 0 : (pSC->m_Dparam->cROITopY & 0xf)), iFirstColumn = pSC->m_Dparam->cROILeftX;
const size_t iAlphaPos = pSC->WMII.cLeadingPadding + (pSC->WMII.cfColorFormat == CMYK ? 4 : 3);//only RGB and CMYK may have interleaved alpha
const PixelI * pA = pSC->m_pNextSC->a0MBbuffer[0];
const U8 nLen = pSC->WMISCP.nLenMantissaOrShift;
const I8 nExpBias = pSC->WMISCP.nExpBias;
size_t iRow, iColumn;
size_t * pOffsetX = pSC->m_Dparam->pOffsetX, * pOffsetY = pSC->m_Dparam->pOffsetY + (pSC->cRow - 1) * 16, iY;
if (CF_RGB != pSC->WMII.cfColorFormat && CMYK != pSC->WMII.cfColorFormat)
return ICERR_ERROR;
if(bd == BD_8){
const PixelI iBias = (1 << (iShift + 7)) + (iShift == 0 ? 0 : (1 << (iShift - 1)));
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
PixelI a = ((pA[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iBias) >> iShift);
((U8 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY)[iAlphaPos] = _CLIP8(a);
}
}
else if(bd == BD_16){
const PixelI iBias = (1 << (iShift + 15)) + (iShift == 0 ? 0 : (1 << (iShift - 1)));
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
PixelI a = (((pA[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iBias) >> iShift) << nLen);
((U16 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY)[iAlphaPos] = _CLIPU16(a);
}
}
else if(bd == BD_16S){
const PixelI iBias = (iShift == 0 ? 0 : (1 << (iShift - 1)));
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
PixelI a = (((pA[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iBias) >> iShift) << nLen);
((I16 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY)[iAlphaPos] = _CLIP16(a);
}
}
else if(bd == BD_16F){
const PixelI iBias = (iShift == 0 ? 0 : (1 << (iShift - 1)));
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
PixelI a = ((pA[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iBias) >> iShift);
((U16 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY)[iAlphaPos] = backwardHalf(a);
}
}
else if(bd == BD_32S){
const PixelI iBias = (iShift == 0 ? 0 : (1 << (iShift - 1)));
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
PixelI a = (((pA[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iBias) >> iShift) << nLen);
((I32 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY)[iAlphaPos] = a;
}
}
else if(bd == BD_32F){
const PixelI iBias = (iShift == 0 ? 0 : (1 << (iShift - 1)));
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
PixelI a = ((pA[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iBias) >> iShift);
((float *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY)[iAlphaPos] = pixel2float (a, nExpBias, nLen);
}
}
else // not supported
return ICERR_ERROR;
}
return ICERR_OK;
}
Int outputMBRow(CWMImageStrCodec * pSC)
{
const COLORFORMAT cfExt = (pSC->m_param.cfColorFormat == Y_ONLY ? Y_ONLY : pSC->WMII.cfColorFormat);
const BITDEPTH_BITS bd = pSC->WMII.bdBitDepth;
const PixelI iShift = (pSC->m_param.bScaledArith ? SHIFTZERO + QPFRACBITS : 0);
const size_t cHeight = min((pSC->m_Dparam->cROIBottomY + 1) - (pSC->cRow - 1) * 16, 16);
const size_t cWidth = (pSC->m_Dparam->cROIRightX + 1);
const size_t iFirstRow = ((pSC->cRow - 1) * 16 > pSC->m_Dparam->cROITopY ? 0 : (pSC->m_Dparam->cROITopY & 0xf)), iFirstColumn = pSC->m_Dparam->cROILeftX;
const PixelI *pY = pSC->a0MBbuffer[0];
const PixelI *pU = (pSC->m_bUVResolutionChange ? pSC->pResU : pSC->a0MBbuffer[1]);
const PixelI *pV = (pSC->m_bUVResolutionChange ? pSC->pResV : pSC->a0MBbuffer[2]);
const PixelI *pA = NULL;
const size_t iB = (pSC->WMII.bRGB ? 2 : 0);
const size_t iR = 2 - iB;
const U8 nLen = pSC->WMISCP.nLenMantissaOrShift;
const I8 nExpBias = pSC->WMISCP.nExpBias;
size_t iRow, iColumn, iIdx;
size_t * pOffsetX = pSC->m_Dparam->pOffsetX, * pOffsetY = pSC->m_Dparam->pOffsetY + (pSC->cRow - 1) * (cfExt == YUV_420 ? 8 : 16), iY;
if (pSC->m_pNextSC) {
assert (pSC->m_param.bScaledArith == pSC->m_pNextSC->m_param.bScaledArith); // will be relaxed later
}
// guard output buffer
if(checkImageBuffer(pSC, pSC->WMII.oOrientation >= O_RCW ? pSC->WMII.cROIHeight : pSC->WMII.cROIWidth, cHeight - iFirstRow) != ICERR_OK)
return ICERR_ERROR;
if(pSC->m_bUVResolutionChange)
interpolateUV(pSC);
if(pSC->WMISCP.bYUVData){
I32 * pDst = (I32 *)pSC->WMIBI.pv + (pSC->cRow - 1) *
(pSC->m_param.cfColorFormat == YUV_420 ? 8 : 16) * pSC->WMIBI.cbStride / sizeof(I32);
switch(pSC->m_param.cfColorFormat){
case Y_ONLY:
case YUV_444:
case NCOMPONENT:
{
PixelI * pChannel[16];
size_t iChannel;
const size_t cChannel = pSC->WMII.cfColorFormat == Y_ONLY ? 1 : pSC->WMISCP.cChannel;
assert(cChannel <= 16);
for(iChannel = 0; iChannel < cChannel; iChannel ++)
pChannel[iChannel & 15] = pSC->a0MBbuffer[iChannel];
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
I32 * pRow = pDst;
for(iColumn = iFirstColumn; iColumn < cWidth; iColumn ++){
for(iChannel = 0; iChannel < cChannel; iChannel ++){
PixelI p = pChannel[iChannel & 15][((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf]];
*pRow++ = p;
}
}
pDst += pSC->WMIBI.cbStride / sizeof(I32);
}
}
break;
case YUV_422:
{
PixelI y0, y1, u, v;
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
I32 * pRow = pDst;
for(iColumn = iFirstColumn; iColumn < cWidth; iColumn += 2){
iIdx = ((iColumn >> 4) << 7) + idxCC[iRow][(iColumn >> 1) & 7];
u = pU[iIdx], v = pV[iIdx];
y0 = pY[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]];
y1 = pY[(((iColumn + 1) >> 4) << 8) + idxCC[iRow][(iColumn + 1) & 15]];
pRow[0] = u, pRow[1] = y0, pRow[2] = v, pRow[3] = y1;
pRow += 4;
}
pDst += pSC->WMIBI.cbStride / sizeof(I32);
}
}
break;
case YUV_420:
{
PixelI y0, y1, y2, y3, u, v;
// const size_t iS4[8][4] = {{0, 1, 2, 3}, {2, 3, 0, 1}, {1, 0, 3, 2}, {3, 2, 1, 0}, {1, 3, 0, 2}, {3, 1, 2, 0}, {0, 2, 1, 3}, {2, 0, 3, 1}};
for(iRow = iFirstRow; iRow < cHeight; iRow += 2){
I32 * pRow = pDst;
for(iColumn = iFirstColumn; iColumn < cWidth; iColumn += 2){
iIdx = ((iColumn >> 4) << 6) + idxCC_420[iRow >> 1][(iColumn >> 1) & 7];
u = pU[iIdx], v = pV[iIdx];
y0 = pY[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]];
y1 = pY[(((iColumn + 1) >> 4) << 8) + idxCC[iRow][(iColumn + 1) & 15]];
y2 = pY[((iColumn >> 4) << 8) + idxCC[iRow + 1][iColumn & 15]];
y3 = pY[(((iColumn + 1) >> 4) << 8) + idxCC[iRow + 1][(iColumn + 1) & 15]];
pRow[0] = y0, pRow[1] = y1, pRow[2] = y2, pRow[3] = y3, pRow[4] = u, pRow[5] = v;
pRow += 6;
}
pDst += pSC->WMIBI.cbStride / sizeof(I32);
}
}
break;
default:
assert(0);
break;
}
}
else if(bd == BD_8){
U8 * pDst;
const PixelI iBias1 = 128 << iShift;
const PixelI iBias2 = pSC->m_param.bScaledArith ? ((1 << (SHIFTZERO + QPFRACBITS - 1)) - 1) : 0;
const PixelI iBias = iBias1 + iBias2;
switch(cfExt){
case CF_RGB:
{
PixelI r, g, b, a;
if (pSC->m_pNextSC && pSC->WMISCP.uAlphaMode > 0) { // RGBA
pA = pSC->m_pNextSC->a0MBbuffer[0];
if (pSC->m_param.bScaledArith == FALSE) {
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
g = pY[iIdx] + iBias, r = -pU[iIdx], b = pV[iIdx];
a = pA[iIdx] + iBias;
_ICC(r, g, b);
pDst = (U8 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
if ((g | b | r | a) & ~0xff)
pDst[iR] = _CLIP8(r), pDst[1] = _CLIP8(g), pDst[iB] = _CLIP8(b), pDst[3] = _CLIP8(a);
else
pDst[iR] = (U8)r, pDst[1] = (U8)g, pDst[iB] = (U8)b, pDst[3] = (U8)(a);
}
}
else{
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
g = pY[iIdx] + iBias, r = -pU[iIdx], b = pV[iIdx];
a = pA[iIdx] + iBias;
_ICC(r, g, b);
g >>= iShift, b >>= iShift, r >>= iShift, a >>= iShift;
pDst = (U8 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
if ((g | b | r | a) & ~0xff)
pDst[iR] = _CLIP8(r), pDst[1] = _CLIP8(g), pDst[iB] = _CLIP8(b), pDst[3] = _CLIP8(a);
else
pDst[iR] = (U8)r, pDst[1] = (U8)g, pDst[iB] = (U8)b, pDst[3] = (U8)(a);
}
}
}
else {
if (pSC->m_param.bScaledArith == FALSE) {
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
g = pY[iIdx] + iBias, r = -pU[iIdx], b = pV[iIdx];
_ICC(r, g, b);
pDst = (U8 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
if ((g | b | r) & ~0xff)
pDst[iR] = _CLIP8(r), pDst[1] = _CLIP8(g), pDst[iB] = _CLIP8(b);
else
pDst[iR] = (U8)r, pDst[1] = (U8)g, pDst[iB] = (U8)b;
}
}
else{
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
g = pY[iIdx] + iBias, r = -pU[iIdx], b = pV[iIdx];
_ICC(r, g, b);
g >>= iShift, b >>= iShift, r >>= iShift;
pDst = (U8 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
if ((g | b | r) & ~0xff)
pDst[iR] = _CLIP8(r), pDst[1] = _CLIP8(g), pDst[iB] = _CLIP8(b);
else
pDst[iR] = (U8)r, pDst[1] = (U8)g, pDst[iB] = (U8)b;
}
}
}
break;
}
case Y_ONLY:
case YUV_444:
case NCOMPONENT:
outputNChannel(pSC, iFirstRow, iFirstColumn, cWidth, cHeight, iShift, iBias);
break;
case YUV_422:
{
PixelI y0, y1, u, v;
// const ORIENTATION oO = pSC->WMII.oOrientation;
// const size_t i0 = ((oO > O_FLIPV && oO <= O_RCW_FLIPVH) ? 1 : 0), i1 = 1 - i0;
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn += 2){
iIdx = ((iColumn >> 4) << 7) + idxCC[iRow][(iColumn >> 1) & 7];
u = ((pU[iIdx] + iBias) >> iShift), v = ((pV[iIdx] + iBias) >> iShift);
y0 = ((pY[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iBias) >> iShift);
y1 = ((pY[(((iColumn + 1) >> 4) << 8) + idxCC[iRow][(iColumn + 1) & 15]] + iBias) >> iShift);
pDst = (U8 *)pSC->WMIBI.pv + pOffsetX[iColumn >> 1] + iY;
if ((y0 | y1 | u | v) & ~0xff)//UYVY
pDst[0] = _CLIP8(u), pDst[1] = _CLIP8(y0), pDst[2] = _CLIP8(v), pDst[3] = _CLIP8(y1);
else
pDst[0] = (U8)u, pDst[1] = (U8)y0, pDst[2] = (U8)v, pDst[3] = (U8)y1;
}
}
}
break;
case YUV_420:
{
PixelI y0, y1, y2, y3, u, v;
const size_t iS4[8][4] = {{0, 1, 2, 3}, {2, 3, 0, 1}, {1, 0, 3, 2}, {3, 2, 1, 0}, {1, 3, 0, 2}, {3, 1, 2, 0}, {0, 2, 1, 3}, {2, 0, 3, 1}};
const ORIENTATION oO = pSC->WMII.oOrientation;
const size_t i0 = iS4[oO][0], i1 = iS4[oO][1], i2 = iS4[oO][2], i3 = iS4[oO][3];
for(iRow = iFirstRow; iRow < cHeight; iRow += 2){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> 1]; iColumn < cWidth; iColumn += 2){
iIdx = ((iColumn >> 4) << 6) + idxCC_420[iRow >> 1][(iColumn >> 1) & 7];
u = ((pU[iIdx] + iBias) >> iShift), v = ((pV[iIdx] + iBias) >> iShift);
y0 = ((pY[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iBias) >> iShift);
y1 = ((pY[(((iColumn + 1) >> 4) << 8) + idxCC[iRow][(iColumn + 1) & 15]] + iBias) >> iShift);
y2 = ((pY[((iColumn >> 4) << 8) + idxCC[iRow + 1][iColumn & 15]] + iBias) >> iShift);
y3 = ((pY[(((iColumn + 1) >> 4) << 8) + idxCC[iRow + 1][(iColumn + 1) & 15]] + iBias) >> iShift);
pDst = (U8 *)pSC->WMIBI.pv + pOffsetX[iColumn >> 1] + iY;
if ((y0 | y1 | y2 | y3 | u | v) & ~0xff)
pDst[i0] = _CLIP8(y0), pDst[i1] = _CLIP8(y1), pDst[i2] = _CLIP8(y2), pDst[i3] = _CLIP8(y3), pDst[4] = _CLIP8(u), pDst[5] = _CLIP8(v);
else
pDst[i0] = (U8)y0, pDst[i1] = (U8)y1, pDst[i2] = (U8)y2, pDst[i3] = (U8)y3, pDst[4] = (U8)u, pDst[5] = (U8)v;
}
}
}
break;
case CMYK:
{
PixelI c, m, y, k;
PixelI * pK = pSC->a0MBbuffer[3];
for(iRow = iFirstRow; iRow < cHeight; iRow++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
m = -pY[iIdx] + iBias1, c = pU[iIdx], y = -pV[iIdx], k = pK[iIdx] + iBias2;
_ICC_CMYK(c, m, y, k); // color conversion
c >>= iShift, m >>= iShift, y >>= iShift, k >>= iShift;
pDst = (U8 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
if ((c | m | y | k) & ~0xff)
pDst[0] = _CLIP8(c), pDst[1] = _CLIP8(m), pDst[2] = _CLIP8(y), pDst[3] = _CLIP8(k);
else
pDst[0] = (U8)c, pDst[1] = (U8)m, pDst[2] = (U8)y, pDst[3] = (U8)k;
}
}
}
break;
case CF_RGBE:
{
PixelI r, g, b;
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
g = pY[iIdx] + iBias2, r = -pU[iIdx], b = pV[iIdx];
_ICC(r, g, b);
pDst = (U8 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
inverseConvertRGBE (r >> iShift, g >> iShift, b >> iShift, pDst, pDst + 1, pDst + 2, pDst + 3);
}
}
}
break;
default:
assert(0);
break;
}
}
else if(bd == BD_16){
const PixelI iBias = (((1 << 15) >> nLen) << iShift) + (iShift == 0 ? 0 : (1 << (iShift - 1)));
U16 * pDst;
switch(cfExt){
case CF_RGB:
{
PixelI r, g, b;
if (pSC->m_param.bScaledArith == FALSE) {
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
g = pY[iIdx] + iBias, r = -pU[iIdx], b = pV[iIdx];
_ICC(r, g, b);
g <<= nLen, b <<= nLen, r <<= nLen;
pDst = (U16 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
if ((g | b | r) & ~0xffff)
pDst[0] = _CLIPU16(r), pDst[1] = _CLIPU16(g), pDst[2] = _CLIPU16(b);
else
pDst[0] = (U16)r, pDst[1] = (U16)g, pDst[2] = (U16)b;
}
}
else{
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
g = pY[iIdx] + iBias, r = -pU[iIdx], b = pV[iIdx];
_ICC(r, g, b);
g = (g >> iShift) << nLen, b = (b >> iShift) << nLen, r = (r >> iShift) << nLen;
pDst = (U16 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
if ((g | b | r) & ~0xffff)
pDst[0] = _CLIPU16(r), pDst[1] = _CLIPU16(g), pDst[2] = _CLIPU16(b);
else
pDst[0] = (U16)r, pDst[1] = (U16)g, pDst[2] = (U16)b;
}
}
break;
}
case Y_ONLY:
case YUV_444:
case NCOMPONENT:
outputNChannel(pSC, iFirstRow, iFirstColumn, cWidth, cHeight, iShift, iBias);
break;
case YUV_422:
{
PixelI y0, y1, u, v;
const ORIENTATION oO = pSC->WMII.oOrientation;
const size_t i0 = ((oO == O_FLIPH || oO == O_FLIPVH || oO == O_RCW_FLIPV || oO == O_RCW_FLIPVH) ? 1 : 0), i1 = 1 - i0;
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn += 2){
iIdx = ((iColumn >> 4) << 7) + idxCC[iRow][(iColumn >> 1) & 7];
u = ((pU[iIdx] + iBias) >> iShift) << nLen, v = ((pV[iIdx] + iBias) >> iShift) << nLen;
y0 = ((pY[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iBias) >> iShift) << nLen;
y1 = ((pY[(((iColumn + 1) >> 4) << 8) + idxCC[iRow][(iColumn + 1) & 15]] + iBias) >> iShift) << nLen;
pDst = (U16 *)pSC->WMIBI.pv + pOffsetX[iColumn >> 1] + iY;
if ((y0 | y1 | u | v) & ~0xffff)
{
pDst[i0] = _CLIPU16(u);
pDst[i1] = _CLIPU16(y0);
pDst[2] = _CLIPU16(v);
pDst[3] = _CLIPU16(y1);
}
else
{
pDst[i0] = (U16)(u);
pDst[i1] = (U16)(y0);
pDst[2] = (U16)(v);
pDst[3] = (U16)(y1);
}
}
}
}
break;
case YUV_420:
{
PixelI y0, y1, y2, y3, u, v;
const size_t iS4[8][4] = {{0, 1, 2, 3}, {2, 3, 0, 1}, {1, 0, 3, 2}, {3, 2, 1, 0}, {1, 3, 0, 2}, {3, 1, 2, 0}, {0, 2, 1, 3}, {2, 0, 3, 1}};
const ORIENTATION oO = pSC->WMII.oOrientation;
const size_t i0 = iS4[oO][0], i1 = iS4[oO][1], i2 = iS4[oO][2], i3 = iS4[oO][3];
for(iRow = iFirstRow; iRow < cHeight; iRow += 2){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> 1]; iColumn < cWidth; iColumn += 2){
iIdx = ((iColumn >> 3) << 6) + idxCC[iRow][(iColumn >> 1) & 7];
u = ((pU[iIdx] + iBias) >> iShift) << nLen, v = ((pV[iIdx] + iBias) >> iShift) << nLen;
y0 = ((pY[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iBias) >> iShift) << nLen;
y1 = ((pY[(((iColumn + 1) >> 4) << 8) + idxCC[iRow][(iColumn + 1) & 15]] + iBias) >> iShift) << nLen;
y2 = ((pY[((iColumn >> 4) << 8) + idxCC[iRow + 1][iColumn & 15]] + iBias) >> iShift) << nLen;
y3 = ((pY[(((iColumn + 1) >> 4) << 8) + idxCC[iRow + 1][(iColumn + 1) & 15]] + iBias) >> iShift) << nLen;
pDst = (U16 *)pSC->WMIBI.pv + pOffsetX[iColumn >> 1] + iY;
if ((y0 | y1 | y2 | y3 | u | v) & ~0xffff)
{
pDst[i0] = _CLIPU16(y0);
pDst[i1] = _CLIPU16(y1);
pDst[i2] = _CLIPU16(y2);
pDst[i3] = _CLIPU16(y3);
pDst[4] = _CLIPU16(u);
pDst[5] = _CLIPU16(v);
}
else
{
pDst[i0] = (U16)(y0);
pDst[i1] = (U16)(y1);
pDst[i2] = (U16)(y2);
pDst[i3] = (U16)(y3);
pDst[4] = (U16)(u);
pDst[5] = (U16)(v);
}
}
}
}
break;
case CMYK:
{
PixelI c, m, y, k;
PixelI * pK = pSC->a0MBbuffer[3];
const PixelI iBias1 = (32768 >> nLen) << iShift;
const PixelI iBias2 = iBias - iBias1;
for(iRow = iFirstRow; iRow < cHeight; iRow++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
m = -pY[iIdx] + iBias1, c = pU[iIdx], y = -pV[iIdx], k = pK[iIdx] + iBias2;
_ICC_CMYK(c, m, y, k); // color conversion
c = (c >> iShift) << nLen, m = (m >> iShift) << nLen, y = (y >> iShift) << nLen, k = (k >> iShift) << nLen;
pDst = (U16 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
if ((c | m | y | k) & ~0xffff)
pDst[0] = _CLIPU16(c), pDst[1] = _CLIPU16(m), pDst[2] = _CLIPU16(y), pDst[3] = _CLIPU16(k);
else
pDst[0] = (U16)(c), pDst[1] = (U16)(m), pDst[2] = (U16)(y), pDst[3] = (U16)(k);
}
}
}
break;
default:
assert(0);
break;
}
}
else if(bd == BD_16S){
const PixelI iBias = pSC->m_param.bScaledArith ? ((1 << (SHIFTZERO + QPFRACBITS - 1)) - 1) : 0;
I16 * pDst;
switch(cfExt){
case CF_RGB:
{
PixelI r, g, b;
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
g = pY[iIdx] + iBias, r = -pU[iIdx], b = pV[iIdx];
_ICC(r, g, b);
r = (r >> iShift) << nLen, g = (g >> iShift) << nLen, b = (b >> iShift) << nLen;
pDst = (I16 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
pDst[0] = _CLIP16(r), pDst[1] = _CLIP16(g), pDst[2] = _CLIP16(b);
}
break;
}
case Y_ONLY:
case YUV_444:
case NCOMPONENT:
outputNChannel(pSC, iFirstRow, iFirstColumn, cWidth, cHeight, iShift, iBias);
break;
case CMYK:
{
PixelI c, m, y, k;
PixelI * pK = pSC->a0MBbuffer[3];
for(iRow = iFirstRow; iRow < cHeight; iRow++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
m = -pY[iIdx], c = pU[iIdx], y = -pV[iIdx], k = pK[iIdx] + iBias;
_ICC_CMYK(c, m, y, k); // color conversion
c = (c >> iShift) << nLen, m = (m >> iShift) << nLen, y = (y >> iShift) << nLen, k = (k >> iShift) << nLen;
pDst = (I16 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
pDst[0] = (I16)(c), pDst[1] = (I16)(m), pDst[2] = (I16)(y), pDst[3] = (I16)(k);
}
}
}
break;
default:
assert(0);
break;
}
}
else if(bd == BD_16F){
const PixelI iBias = pSC->m_param.bScaledArith ? ((1 << (SHIFTZERO + QPFRACBITS - 1)) - 1) : 0;
U16 *pDst;
switch (cfExt)
{
case CF_RGB:
{
PixelI r, g, b;
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
g = pY[iIdx] + iBias, r = -pU[iIdx], b = pV[iIdx];
_ICC(r, g, b);
pDst = (U16 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
pDst[0] = backwardHalf(r >> iShift);
pDst[1] = backwardHalf(g >> iShift);
pDst[2] = backwardHalf(b >> iShift);
}
}
break;
}
case Y_ONLY:
case YUV_444:
case NCOMPONENT:
outputNChannel(pSC, iFirstRow, iFirstColumn, cWidth, cHeight, iShift, iBias);
break;
default:
assert(0);
break;
}
}
else if(bd == BD_32){
const PixelI iBias = (((1 << 31) >> nLen) << iShift) + (iShift == 0 ? 0 : (1 << (iShift - 1)));
U32 * pDst;
switch (cfExt)
{
case CF_RGB:
{
PixelI r, g, b;
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
g = pY[iIdx] + iBias, r = -pU[iIdx], b = pV[iIdx];
_ICC(r, g, b);
pDst = (U32 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
pDst[0] = ((r >> iShift) << nLen);
pDst[1] = ((g >> iShift) << nLen);
pDst[2] = ((b >> iShift) << nLen);
}
}
}
break;
case Y_ONLY:
case YUV_444:
case NCOMPONENT:
{
outputNChannel(pSC, iFirstRow, iFirstColumn, cWidth, cHeight, iShift, iBias);
break;
}
default:
assert(0);
break;
}
}
else if(bd == BD_32S){
const PixelI iBias = pSC->m_param.bScaledArith ? ((1 << (SHIFTZERO + QPFRACBITS - 1)) - 1) : 0;
int * pDst;
switch (cfExt)
{
case CF_RGB:
{
PixelI r, g, b;
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
g = pY[iIdx] + iBias, r = -pU[iIdx], b = pV[iIdx];
_ICC(r, g, b);
pDst = (int *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
pDst[0] = ((r >> iShift) << nLen);
pDst[1] = ((g >> iShift) << nLen);
pDst[2] = ((b >> iShift) << nLen);
}
}
break;
}
case Y_ONLY:
case YUV_444:
case NCOMPONENT:
outputNChannel(pSC, iFirstRow, iFirstColumn, cWidth, cHeight, iShift, iBias);
break;
default:
assert(0);
break;
}
}
else if(bd == BD_32F){
const PixelI iBias = pSC->m_param.bScaledArith ? ((1 << (SHIFTZERO + QPFRACBITS - 1)) - 1) : 0;
float * pDst;
switch (cfExt)
{
case CF_RGB:
{
PixelI r, g, b;
for(iRow = iFirstRow; iRow < cHeight; iRow ++){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
g = pY[iIdx] + iBias, r = -pU[iIdx], b = pV[iIdx];
_ICC(r, g, b);
pDst = (float *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
pDst[0] = pixel2float (r >> iShift, nExpBias, nLen);
pDst[1] = pixel2float (g >> iShift, nExpBias, nLen);
pDst[2] = pixel2float (b >> iShift, nExpBias, nLen);
}
}
break;
}
case Y_ONLY:
case YUV_444:
case NCOMPONENT:
outputNChannel(pSC, iFirstRow, iFirstColumn, cWidth, cHeight, iShift, iBias);
break;
default:
assert(0);
break;
}
}
else if(bd == BD_5){
const PixelI iBias = (16 << iShift) + (pSC->m_param.bScaledArith ? ((1 << (SHIFTZERO + QPFRACBITS - 1)) - 1) : 0);
PixelI r, g, b;
U16 * pDst;
assert(cfExt == CF_RGB);
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
g = pY[iIdx] + iBias, r = -pU[iIdx], b = pV[iIdx];
_ICC(r, g, b);
g >>= iShift, b >>= iShift, r >>= iShift;
pDst = (U16 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
if (pSC->m_param.bRBSwapped)
pDst[0] = (U16)_CLIP2(0, b, 31) + (((U16)_CLIP2(0, g, 31)) << 5) + (((U16)_CLIP2(0, r, 31)) << 10);
else
pDst[0] = (U16)_CLIP2(0, r, 31) + (((U16)_CLIP2(0, g, 31)) << 5) + (((U16)_CLIP2(0, b, 31)) << 10);
}
}
else if(bd == BD_565){
const PixelI iBias = (32 << iShift) + (pSC->m_param.bScaledArith ? ((1 << (SHIFTZERO + QPFRACBITS - 1)) - 1) : 0);
PixelI r, g, b;
U16 * pDst;
assert(cfExt == CF_RGB);
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
g = pY[iIdx] + iBias, r = -pU[iIdx], b = pV[iIdx];
_ICC(r, g, b);
g >>= iShift, b >>= iShift + 1, r >>= iShift + 1;
pDst = (U16 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
if (pSC->m_param.bRBSwapped)
pDst[0] = (U16)_CLIP2(0, b, 31) + (((U16)_CLIP2(0, g, 63)) << 5) + (((U16)_CLIP2(0, r, 31)) << 11);
else
pDst[0] = (U16)_CLIP2(0, r, 31) + (((U16)_CLIP2(0, g, 63)) << 5) + (((U16)_CLIP2(0, b, 31)) << 11);
}
}
else if(bd == BD_10){
const PixelI iBias = (512 << iShift) + (pSC->m_param.bScaledArith ? ((1 << (SHIFTZERO + QPFRACBITS - 1)) - 1) : 0);
PixelI r, g, b;
U32 * pDst;
assert(cfExt == CF_RGB);
for(iRow = iFirstRow; iRow < cHeight; iRow ++)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){
iIdx = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
g = pY[iIdx] + iBias, r = -pU[iIdx], b = pV[iIdx];
_ICC(r, g, b);
g >>= iShift, b >>= iShift, r >>= iShift;
pDst = (U32 *)pSC->WMIBI.pv + pOffsetX[iColumn] + iY;
if (pSC->m_param.bRBSwapped)
pDst[0] = (U32)_CLIP2(0, b, 1023) +
(((U32)_CLIP2(0, g, 1023)) << 10) +
(((U32)_CLIP2(0, r, 1023)) << 20);
else
pDst[0] = (U32)_CLIP2(0, r, 1023) +
(((U32)_CLIP2(0, g, 1023)) << 10) +
(((U32)_CLIP2(0, b, 1023)) << 20);
}
}
else if(bd == BD_1){
const size_t iPos = pSC->WMII.cLeadingPadding;
const Int iTh = (iShift > 0) ? (1 << (iShift - 1)) : 1;
assert(cfExt == Y_ONLY && pSC->m_param.cfColorFormat == Y_ONLY);
if(pSC->WMII.oOrientation < O_RCW)
for(iRow = iFirstRow; iRow < cHeight; iRow ++) {
iY = pOffsetY[iRow] + iPos;
for(iColumn = iFirstColumn; iColumn < cWidth; iColumn ++) {
U8 cByte = ((U8 *)pSC->WMIBI.pv + (pOffsetX[iColumn] >> 3) + iY)[0];
U8 cShift = (U8)(7 - (pOffsetX[iColumn] & 7));
((U8 *)pSC->WMIBI.pv + (pOffsetX[iColumn] >> 3) + iY)[0] ^= // exor is used because we can't assume the byte was originally zero
(((pSC->WMISCP.bBlackWhite + (pY[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] >= iTh)
+ (cByte >> cShift)) & 0x1) << cShift);
}
}
else
for(iRow = iFirstRow; iRow < cHeight; iRow ++) {
iY = pOffsetY[iRow] + iPos;
for(iColumn = iFirstColumn; iColumn < cWidth; iColumn ++) {
U8 cByte = ((U8 *)pSC->WMIBI.pv + pOffsetX[iColumn] + (iY >> 3))[0];
U8 cShift = (U8)(7 - (iY & 7)); // should be optimized out
((U8 *)pSC->WMIBI.pv + pOffsetX[iColumn] + (iY >> 3))[0] ^= // exor is used because we can't assume the byte was originally zero
(((pSC->WMISCP.bBlackWhite + (pY[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] >= iTh)
+ (cByte >> cShift)) & 0x1) << cShift);
}
}
}
if(pSC->WMISCP.uAlphaMode > 0)
if(outputMBRowAlpha(pSC) != ICERR_OK)
return ICERR_ERROR;
#ifdef REENTRANT_MODE
pSC->WMIBI.cLinesDecoded = cHeight - iFirstRow;
if (CF_RGB == pSC->WMII.cfColorFormat && Y_ONLY == pSC->WMISCP.cfColorFormat)
{
const CWMImageInfo* pII = &pSC->WMII;
#define fixupFullSize(type, nCh) \
for(iRow = iFirstRow; iRow < cHeight; iRow ++) {\
size_t iY;\
for(iColumn = iFirstColumn, iY = pOffsetY[iRow]; iColumn < cWidth; iColumn ++){\
type *pT = (type*)(U8 *)pSC->WMIBI.pv + iY + pOffsetX[iColumn];\
pT[2] = pT[1] = pT[0]; \
pT += nCh; \
} \
} \
break
switch (pII->bdBitDepth)
{
case BD_8:
fixupFullSize(U8, (pII->cBitsPerUnit >> 3));
break;
case BD_16:
case BD_16S:
case BD_16F:
fixupFullSize(U16, (pII->cBitsPerUnit >> 3) / sizeof(U16));
break;
case BD_32:
case BD_32S:
case BD_32F:
fixupFullSize(U32, (pII->cBitsPerUnit >> 3) / sizeof(float));
break;
case BD_5:
case BD_10:
case BD_565:
default:
break;
}
}
#endif
return ICERR_OK;
}
// Y_ONLY/CF_ALPHA/YUV_444/N_CHANNEL thumbnail decode
Void outputNChannelThumbnail(CWMImageStrCodec * pSC, const PixelI cMul, const size_t rShiftY, size_t iFirstRow, size_t iFirstColumn)
{
const size_t tScale = pSC->m_Dparam->cThumbnailScale;
const size_t cWidth = (pSC->m_Dparam->cROIRightX + 1);
const size_t cHeight = min((pSC->m_Dparam->cROIBottomY + 1) - (pSC->cRow - 1) * 16, 16);
const size_t cChannel = pSC->WMISCP.cChannel;
const U8 nLen = pSC->WMISCP.nLenMantissaOrShift;
const I8 nExpBias = pSC->WMISCP.nExpBias;
size_t nBits = 0;
PixelI iOffset;
PixelI * pChannel[16];
size_t iChannel, iRow, iColumn;
size_t * pOffsetX = pSC->m_Dparam->pOffsetX, * pOffsetY = pSC->m_Dparam->pOffsetY + (pSC->cRow - 1) * 16 / tScale, iY;
while((size_t)(1U << nBits) < tScale)
nBits ++;
assert(cChannel <= 16);
for(iChannel = 0; iChannel < cChannel; iChannel ++)
pChannel[iChannel & 15] = pSC->a0MBbuffer[iChannel];
if(pSC->m_bUVResolutionChange)
pChannel[1] = pSC->pResU, pChannel[2] = pSC->pResV;
switch(pSC->WMII.bdBitDepth){
case BD_8:
for(iOffset = (128 << rShiftY) / cMul, iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
U8 * pDst = (U8 *)pSC->WMIBI.pv + iY + pOffsetX[iColumn >> nBits];
for(iChannel = 0; iChannel < cChannel; iChannel ++){
PixelI p = ((pChannel[iChannel & 15][((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iOffset) * cMul) >> rShiftY;
pDst[iChannel] = _CLIP8(p);
}
}
}
break;
case BD_16:
for(iOffset = (32768 << rShiftY) / cMul, iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
U16 * pDst = (U16 *)pSC->WMIBI.pv + iY + pOffsetX[iColumn >> nBits];
for(iChannel = 0; iChannel < cChannel; iChannel ++){
PixelI p = (((pChannel[iChannel & 15][((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iOffset) * cMul) >> rShiftY) << nLen;
pDst[iChannel] = _CLIPU16(p);
}
}
}
break;
case BD_16S:
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
I16 * pDst = (I16 *)pSC->WMIBI.pv + iY + pOffsetX[iColumn >> nBits];
for(iChannel = 0; iChannel < cChannel; iChannel ++){
PixelI p = ((pChannel[iChannel & 15][((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] * cMul) >> rShiftY) << nLen;
pDst[iChannel] = _CLIP16(p);
}
}
}
break;
case BD_16F:
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
U16 * pDst = (U16 *)pSC->WMIBI.pv + iY + pOffsetX[iColumn >> nBits];
for(iChannel = 0; iChannel < cChannel; iChannel ++){
PixelI p = (pChannel[iChannel & 15][((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] * cMul) >> rShiftY;
pDst[iChannel] = backwardHalf(p);
}
}
}
break;
case BD_32:
for(iOffset = (((1 << 31) >> nLen) << rShiftY) / cMul, iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
U32 * pDst = (U32 *)pSC->WMIBI.pv + iY + pOffsetX[iColumn >> nBits];
for(iChannel = 0; iChannel < cChannel; iChannel ++){
PixelI p = (((pChannel[iChannel & 15][((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] + iOffset) * cMul) >> rShiftY) << nLen;
pDst[iChannel] = (U32)(p);
}
}
}
break;
case BD_32S:
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
I32 * pDst = (I32 *)pSC->WMIBI.pv + iY + pOffsetX[iColumn >> nBits];
for(iChannel = 0; iChannel < cChannel; iChannel ++){
PixelI p = ((pChannel[iChannel & 15][((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] * cMul) >> rShiftY) << nLen;
pDst[iChannel] = (I32)(p);
}
}
}
break;
case BD_32F:
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
float * pDst = (float *)pSC->WMIBI.pv + iY + pOffsetX[iColumn >> nBits];
for(iChannel = 0; iChannel < cChannel; iChannel ++){
PixelI p = (pChannel[iChannel & 15][((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] * cMul) >> rShiftY;
pDst[iChannel] = pixel2float (p, nExpBias, nLen);
}
}
}
break;
default:
assert(0);
break;
}
}
// centralized alpha channel thumbnail, small perf penalty
Int decodeThumbnailAlpha(CWMImageStrCodec * pSC, const size_t nBits, const PixelI cMul, const size_t rShiftY)
{
if(pSC->m_bSecondary == FALSE && pSC->m_pNextSC != NULL){ // with alpha channel
const size_t tScale = (size_t)(1U << nBits);
const size_t cHeight = min((pSC->m_Dparam->cROIBottomY + 1) - (pSC->cRow - 1) * 16, 16);
const size_t cWidth = (pSC->m_Dparam->cROIRightX + 1);
const size_t iFirstRow = ((((pSC->cRow - 1) * 16 > pSC->m_Dparam->cROITopY ? 0 : (pSC->m_Dparam->cROITopY & 0xf)) + tScale - 1) / tScale * tScale);
const size_t iFirstColumn = (pSC->m_Dparam->cROILeftX + tScale - 1) / tScale * tScale;
const size_t iAlphaPos = pSC->WMII.cLeadingPadding + (pSC->WMII.cfColorFormat == CMYK ? 4 : 3);//only RGB and CMYK may have interleaved alpha
const BITDEPTH_BITS bd = pSC->WMII.bdBitDepth;
const PixelI * pSrc = pSC->m_pNextSC->a0MBbuffer[0];
const U8 nLen = pSC->m_pNextSC->WMISCP.nLenMantissaOrShift;
const I8 nExpBias = pSC->m_pNextSC->WMISCP.nExpBias;
size_t iRow, iColumn;
size_t * pOffsetX = pSC->m_Dparam->pOffsetX, * pOffsetY = pSC->m_Dparam->pOffsetY + (pSC->cRow - 1) * 16 / tScale, iY;
if (CF_RGB != pSC->WMII.cfColorFormat && CMYK != pSC->WMII.cfColorFormat)
return ICERR_ERROR;
if(bd == BD_8){
const PixelI offset = (128 << rShiftY) / cMul;
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
PixelI a = ((pSrc[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf]] + offset) * cMul) >> rShiftY;
((U8 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY)[iAlphaPos] = _CLIP8(a);
}
}
else if(bd == BD_16){
const PixelI offset = (32768 << rShiftY) / cMul;
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
PixelI a = (((pSrc[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf]] + offset) * cMul) >> rShiftY) << nLen;
((U16 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY)[iAlphaPos] = _CLIPU16(a);
}
}
else if(bd == BD_16S){
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
PixelI a = ((pSrc[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf]] * cMul) >> rShiftY) << nLen;
((I16 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY)[iAlphaPos] = _CLIP16(a);
}
}
else if(bd == BD_16F){
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
PixelI a = (pSrc[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf]] * cMul) >> rShiftY;
((U16 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY)[iAlphaPos] = backwardHalf(a);
}
}
else if(bd == BD_32S){
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
PixelI a = ((pSrc[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf]] * cMul) >> rShiftY) << nLen;
((I32 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY)[iAlphaPos] = a;
}
}
else if(bd == BD_32F){
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
PixelI a = (pSrc[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf]] * cMul) >> rShiftY;
((float *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY)[iAlphaPos] = pixel2float (a, nExpBias, nLen);
}
}
else // not supported
return ICERR_ERROR;
}
return ICERR_OK;
}
Int decodeThumbnail(CWMImageStrCodec * pSC)
{
const size_t tScale = pSC->m_Dparam->cThumbnailScale;
const size_t cHeight = min((pSC->m_Dparam->bDecodeFullFrame ? pSC->WMII.cHeight : pSC->m_Dparam->cROIBottomY + 1) - (pSC->cRow - 1) * 16, 16);
const size_t cWidth = (pSC->m_Dparam->bDecodeFullFrame ? pSC->WMII.cWidth : pSC->m_Dparam->cROIRightX + 1);
const size_t iFirstRow = ((((pSC->cRow - 1) * 16 > pSC->m_Dparam->cROITopY ? 0 : (pSC->m_Dparam->cROITopY & 0xf)) + tScale - 1) / tScale * tScale);
const size_t iFirstColumn = (pSC->m_Dparam->cROILeftX + tScale - 1) / tScale * tScale;
const COLORFORMAT cfInt = pSC->m_param.cfColorFormat;
const COLORFORMAT cfExt = (pSC->m_param.cfColorFormat == Y_ONLY ? Y_ONLY : pSC->WMII.cfColorFormat);
const BITDEPTH_BITS bd = pSC->WMII.bdBitDepth;
const OVERLAP ol = pSC->WMISCP.olOverlap;
const size_t iB = (pSC->WMII.bRGB ? 2 : 0);
const size_t iR = 2 - iB;
const U8 nLen = pSC->WMISCP.nLenMantissaOrShift;
const I8 nExpBias = pSC->WMISCP.nExpBias;
PixelI offset;
size_t iRow, iColumn, iIdx1, iIdx2, iIdx3 = 0, nBits = 0;
PixelI * pSrcY = pSC->a0MBbuffer[0];
PixelI * pSrcU = pSC->a0MBbuffer[1], * pSrcV = pSC->a0MBbuffer[2];
size_t * pOffsetX = pSC->m_Dparam->pOffsetX, * pOffsetY = pSC->m_Dparam->pOffsetY + (pSC->cRow - 1) * 16 / tScale, iY;
const PixelI cMul = (tScale >= 16 ? (ol == OL_NONE ? 16 : (ol == OL_ONE ? 23 : 34)) : (tScale >= 4 ? (ol == OL_NONE ? 64 : 93) : 258));
const size_t rShiftY = 8 + (pSC->m_param.bScaledArith ? (SHIFTZERO + QPFRACBITS) : 0);
const size_t rShiftUV = rShiftY - ((pSC->m_param.bScaledArith && tScale >= 16) ? ((cfInt == YUV_420 || cfInt == YUV_422) ? 2 : 1) : 0);
while((size_t)(1U << nBits) < tScale)
nBits ++;
assert(tScale == (size_t)(1U << nBits));
// guard output buffer
if(checkImageBuffer(pSC, pSC->WMII.oOrientation < O_RCW ? pSC->WMII.cROIWidth : pSC->WMII.cROIHeight, (cHeight - iFirstRow) / pSC->m_Dparam->cThumbnailScale) != ICERR_OK)
return ICERR_ERROR;
if((((pSC->cRow - 1) * 16) % tScale) != 0)
return ICERR_OK;
if(pSC->cRow * 16 <= pSC->m_Dparam->cROITopY || pSC->cRow * 16 > pSC->m_Dparam->cROIBottomY + 16)
return ICERR_OK;
if((cfInt == YUV_422 || cfInt == YUV_420) && cfExt != Y_ONLY){
PixelI * pDstU = pSC->pResU, * pDstV = pSC->pResV;
for(iRow = 0; iRow < 16; iRow += tScale){
for(iColumn = 0; iColumn < cWidth; iColumn += tScale){
iIdx1 = (cfInt == YUV_422 ? ((iColumn >> 4) << 7) + idxCC[iRow][(iColumn >> 1) & 7] : ((iColumn >> 4) << 6) + idxCC_420[iRow >> 1][(iColumn >> 1) & 7]);
iIdx2 = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
// copy over
pDstU[iIdx2] = pSrcU[iIdx1];
pDstV[iIdx2] = pSrcV[iIdx1];
}
}
if(tScale == 4){
if(cfInt == YUV_420){
for(iColumn = 0; iColumn < cWidth; iColumn += 8){
iIdx1 = ((iColumn >> 4) << 8) + idxCC[0][iColumn & 15];
iIdx2 = ((iColumn >> 4) << 8) + idxCC[4][iColumn & 15];
iIdx3 = ((iColumn >> 4) << 8) + idxCC[8][iColumn & 15];
pDstU[iIdx2] = ((pDstU[iIdx1] + pDstU[iIdx3] + 1) >> 1);
pDstV[iIdx2] = ((pDstV[iIdx1] + pDstV[iIdx3] + 1) >> 1);
iIdx1 = ((iColumn >> 4) << 8) + idxCC[12][iColumn & 15];
pDstU[iIdx1] = pDstU[iIdx3];
pDstV[iIdx1] = pDstV[iIdx3];
}
}
for(iRow = 0; iRow < 16; iRow += 4){
for(iColumn = 0; iColumn < cWidth - 8; iColumn += 8){
iIdx1 = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15];
iIdx2 = ((iColumn >> 4) << 8) + idxCC[iRow][(iColumn + 4) & 15];
iIdx3 = ((iColumn >> 4) << 8) + idxCC[iRow][(iColumn + 8) & 15];
pDstU[iIdx2] = ((pDstU[iIdx1] + pDstU[iIdx3] + 1) >> 1);
pDstV[iIdx2] = ((pDstV[iIdx1] + pDstV[iIdx3] + 1) >> 1);
}
iIdx2 = ((iColumn >> 4) << 8) + idxCC[iRow][(iColumn + 4) & 15];
pDstU[iIdx2] = pDstU[iIdx3];
pDstV[iIdx2] = pDstV[iIdx3];
}
}
pSrcU = pDstU, pSrcV = pDstV;
}
if(bd == BD_8){
U8 * pDst;
offset = (128 << rShiftY) / cMul;
switch(cfExt){
case CF_RGB:
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
size_t iPos = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf];
PixelI g = ((pSrcY[iPos] + offset) * cMul) >> rShiftY, r = -(pSrcU[iPos] * cMul) >> rShiftUV, b = (pSrcV[iPos] * cMul) >> rShiftUV;
_ICC(r, g, b);
pDst = (U8 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY;
pDst[iB] = _CLIP8(b), pDst[1] = _CLIP8(g), pDst[iR] = _CLIP8(r);
}
}
break;
case Y_ONLY:
case YUV_444:
case NCOMPONENT:
outputNChannelThumbnail(pSC, cMul, rShiftY, iFirstRow, iFirstColumn);
break;
case CF_RGBE:
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
size_t iPos = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf];
PixelI g = ((pSrcY[iPos] * cMul) >> rShiftY), r = - ((pSrcU[iPos] * cMul) >> rShiftUV), b = ((pSrcV[iPos] * cMul) >> rShiftUV);
_ICC(r, g, b);
pDst = (U8 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY;
inverseConvertRGBE (r, g, b, pDst, pDst + 1, pDst + 2, pDst + 3);
}
}
break;
case CMYK:
{
PixelI * pSrcK = pSC->a0MBbuffer[3];
PixelI iBias1 = (128 << rShiftY) / cMul, iBias2 = (((128 << rShiftUV) / cMul) >> 1);
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
size_t iPos = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf];
PixelI m = ((-pSrcY[iPos] + iBias1) * cMul) >> rShiftY, c = (pSrcU[iPos] * cMul) >> rShiftUV, y = -(pSrcV[iPos] * cMul) >> rShiftUV, k = ((pSrcK[iPos] + iBias2) * cMul) >> rShiftUV;
_ICC_CMYK(c, m, y, k);
pDst = (U8 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY;
pDst[0] = _CLIP8(c), pDst[1] = _CLIP8(m), pDst[2] = _CLIP8(y), pDst[3] = _CLIP8(k);
}
}
break;
}
default:
assert(0);
break;
}
}
if(bd == BD_16){
U16 * pDst;
offset = (((1 << 15) >> nLen) << rShiftY) / cMul;
switch(cfExt){
case CF_RGB:
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
size_t iPos = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf];
PixelI g = ((pSrcY[iPos] + offset) * cMul) >> rShiftY, r = -(pSrcU[iPos] * cMul) >> rShiftUV, b = (pSrcV[iPos] * cMul) >> rShiftUV;
_ICC(r, g, b);
pDst = (U16 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY;
r <<= nLen, g <<= nLen, b <<= nLen;
pDst[0] = _CLIPU16(r);
pDst[1] = _CLIPU16(g);
pDst[2] = _CLIPU16(b);
}
}
break;
case Y_ONLY:
case YUV_444:
case NCOMPONENT:
outputNChannelThumbnail(pSC, cMul, rShiftY, iFirstRow, iFirstColumn);
break;
case CMYK:
{
PixelI * pSrcK = pSC->a0MBbuffer[3];
PixelI iBias1 = (32768 << rShiftY) / cMul, iBias2 = (((32768 << rShiftUV) / cMul) >> 1);
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
size_t iPos = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf];
PixelI m = ((-pSrcY[iPos] + iBias1) * cMul) >> rShiftY, c = (pSrcU[iPos] * cMul) >> rShiftUV, y = -(pSrcV[iPos] * cMul) >> rShiftUV, k = ((pSrcK[iPos] + iBias2) * cMul) >> rShiftUV;
_ICC_CMYK(c, m, y, k);
pDst = (U16 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY;
c <<= nLen, m <<= nLen, y <<= nLen, k <<= nLen;
pDst[0] = _CLIPU16(c);
pDst[1] = _CLIPU16(m);
pDst[2] = _CLIPU16(y);
pDst[3] = _CLIPU16(k);
}
}
break;
}
default:
assert(0);
break;
}
}
if(bd == BD_16S){
I16 * pDst;
switch(cfExt){
case CF_RGB:
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
size_t iPos = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf];
PixelI g = (pSrcY[iPos] * cMul) >> rShiftY, r = -(pSrcU[iPos] * cMul) >> rShiftUV, b = (pSrcV[iPos] * cMul) >> rShiftUV;
_ICC(r, g, b);
pDst = (I16 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY;
r <<= nLen, g <<= nLen, b <<= nLen;
pDst[0] = _CLIP16(r);
pDst[1] = _CLIP16(g);
pDst[2] = _CLIP16(b);
}
}
break;
case Y_ONLY:
case YUV_444:
case NCOMPONENT:
outputNChannelThumbnail(pSC, cMul, rShiftY, iFirstRow, iFirstColumn);
break;
case CMYK:
{
PixelI * pSrcK = pSC->a0MBbuffer[3];
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
size_t iPos = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf];
PixelI m = -(pSrcY[iPos] * cMul) >> rShiftY, c = (pSrcU[iPos] * cMul) >> rShiftUV, y = -(pSrcV[iPos] * cMul) >> rShiftUV, k = (pSrcK[iPos] * cMul) >> rShiftUV;
_ICC_CMYK(c, m, y, k);
pDst = (I16 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY;
c <<= nLen, m <<= nLen, y <<= nLen, k <<= nLen;
pDst[0] = _CLIP16(c);
pDst[1] = _CLIP16(m);
pDst[2] = _CLIP16(y);
pDst[3] = _CLIP16(k);
}
}
}
break;
default:
assert(0);
break;
}
}
else if(bd == BD_16F){
U16 * pDst;
switch(cfExt){
case CF_RGB:
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
size_t iPos = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf];
PixelI g = (pSrcY[iPos] * cMul) >> rShiftY, r = -(pSrcU[iPos] * cMul) >> rShiftUV, b = (pSrcV[iPos] * cMul) >> rShiftUV;
_ICC(r, g, b);
pDst = (U16 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY;
pDst[0] = backwardHalf (r);
pDst[1] = backwardHalf (g);
pDst[2] = backwardHalf (b);
}
}
break;
case Y_ONLY:
case YUV_444:
case NCOMPONENT:
outputNChannelThumbnail(pSC, cMul, rShiftY, iFirstRow, iFirstColumn);
break;
default:
assert(0);
break;
}
}
else if(bd == BD_32){
U32 * pDst;
offset = (((1 << 31) >> nLen) << rShiftY) / cMul;
switch(cfExt){
case CF_RGB:
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
size_t iPos = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf];
PixelI g = ((pSrcY[iPos] + offset) * cMul) >> rShiftY, r = -(pSrcU[iPos] * cMul) >> rShiftUV, b = (pSrcV[iPos] * cMul) >> rShiftUV;
_ICC(r, g, b);
pDst = (U32 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY;
pDst[0] = (U32)(r << nLen);
pDst[1] = (U32)(g << nLen);
pDst[2] = (U32)(b << nLen);
}
}
break;
case Y_ONLY:
case YUV_444:
case NCOMPONENT:
outputNChannelThumbnail(pSC, cMul, rShiftY, iFirstRow, iFirstColumn);
break;
default:
assert(0);
break;
}
}
else if(bd == BD_32S){
I32 * pDst;
switch(cfExt){
case CF_RGB:
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
size_t iPos = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf];
PixelI g = (pSrcY[iPos] * cMul) >> rShiftY, r = -(pSrcU[iPos] * cMul) >> rShiftUV, b = (pSrcV[iPos] * cMul) >> rShiftUV;
_ICC(r, g, b);
pDst = (I32 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY;
pDst[0] = (I32)(r << nLen);
pDst[1] = (I32)(g << nLen);
pDst[2] = (I32)(b << nLen);
}
}
break;
case Y_ONLY:
case YUV_444:
case NCOMPONENT:
outputNChannelThumbnail(pSC, cMul, rShiftY, iFirstRow, iFirstColumn);
break;
default:
assert(0);
break;
}
}
else if(bd == BD_32F){
float * pDst;
switch(cfExt){
case CF_RGB:
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
size_t iPos = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf];
PixelI g = (pSrcY[iPos] * cMul) >> rShiftY, r = -(pSrcU[iPos] * cMul) >> rShiftUV, b = (pSrcV[iPos] * cMul) >> rShiftUV;
_ICC(r, g, b);
pDst = (float *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY;
pDst[0] = pixel2float (r, nExpBias, nLen);
pDst[1] = pixel2float (g, nExpBias, nLen);
pDst[2] = pixel2float (b, nExpBias, nLen);
}
}
break;
case Y_ONLY:
case YUV_444:
case NCOMPONENT:
outputNChannelThumbnail(pSC, cMul, rShiftY, iFirstRow, iFirstColumn);
break;
default:
assert(0);
break;
}
}
else if(bd == BD_1){
const size_t iPos = pSC->WMII.cLeadingPadding;
Bool bBW;
U8 cByte, cShift;
assert(cfExt == Y_ONLY && pSC->m_param.cfColorFormat == Y_ONLY);
if(pSC->WMII.oOrientation < O_RCW){
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits] + iPos; iColumn < cWidth; iColumn += tScale){
bBW = (pSC->WMISCP.bBlackWhite ^ (pSrcY[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] > 0));
cByte = ((U8 *)pSC->WMIBI.pv + (pOffsetX[iColumn >> nBits] >> 3) + iY)[0];
cShift = (U8)(7 - (pOffsetX[iColumn >> nBits] & 7));
((U8 *)pSC->WMIBI.pv + (pOffsetX[iColumn >> nBits] >> 3) + iY)[0] ^= ((((bBW + (cByte >> cShift)) & 0x1)) << cShift);
}
}
else{
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale)
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits] + iPos; iColumn < cWidth; iColumn += tScale){
bBW = (pSC->WMISCP.bBlackWhite ^ (pSrcY[((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 15]] > 0));
cByte = ((U8 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + (iY >> 3))[0];
cShift = (U8)(7 - (iY & 7));
((U8 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + (iY >> 3))[0] ^= ((((bBW + (cByte >> cShift)) & 0x1)) << cShift);
}
}
}
else if(bd == BD_5){
U16 * pDst;
offset = (16 << rShiftY) / cMul;
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
size_t iPos = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf];
PixelI g = ((pSrcY[iPos] + offset) * cMul) >> rShiftY, r = -(pSrcU[iPos] * cMul) >> rShiftUV, b = (pSrcV[iPos] * cMul) >> rShiftUV;
_ICC(r, g, b);
pDst = (U16 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY;
pDst[0] = (U16)_CLIP2(0, r, 31) + (((U16)_CLIP2(0, g, 31)) << 5) + (((U16)_CLIP2(0, b, 31)) << 10);
}
}
}
else if(bd == BD_565){
U16 * pDst;
offset = (32 << rShiftY) / cMul;
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
size_t iPos = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf];
PixelI g = ((pSrcY[iPos] + offset) * cMul) >> rShiftY, r = -(pSrcU[iPos] * cMul) >> rShiftUV, b = (pSrcV[iPos] * cMul) >> rShiftUV;
_ICC(r, g, b);
r /= 2, b /= 2;
pDst = (U16 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY;
pDst[0] = (U16)_CLIP2(0, r, 31) + (((U16)_CLIP2(0, g, 63)) << 5) + (((U16)_CLIP2(0, b, 31)) << 11);
}
}
}
else if(bd == BD_10){
U32 * pDst;
offset = (512 << rShiftY) / cMul;
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale){
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){
size_t iPos = ((iColumn >> 4) << 8) + idxCC[iRow][iColumn & 0xf];
PixelI g = ((pSrcY[iPos] + offset) * cMul) >> rShiftY, r = -(pSrcU[iPos] * cMul) >> rShiftUV, b = (pSrcV[iPos] * cMul) >> rShiftUV;
_ICC(r, g, b);
pDst = (U32 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY;
pDst[0] = (U32)_CLIP2(0, r, 1023) +
(((U32)_CLIP2(0, g, 1023)) << 10) +
(((U32)_CLIP2(0, b, 1023)) << 20);
}
}
}
if(pSC->WMISCP.uAlphaMode > 0)
if(decodeThumbnailAlpha(pSC, nBits, cMul, rShiftY) != ICERR_OK)
return ICERR_ERROR;
#ifdef REENTRANT_MODE
pSC->WMIBI.cLinesDecoded = ( cHeight - iFirstRow + tScale - 1 ) / tScale;
if (CF_RGB == pSC->WMII.cfColorFormat && Y_ONLY == pSC->WMISCP.cfColorFormat)
{
const CWMImageInfo* pII = &pSC->WMII;
#define fixupThumb(type, nCh) \
for(iRow = iFirstRow; iRow < cHeight; iRow += tScale) {\
size_t iY;\
for(iColumn = iFirstColumn, iY = pOffsetY[iRow >> nBits]; iColumn < cWidth; iColumn += tScale){\
type *pT = (type*)((U8 *)pSC->WMIBI.pv + pOffsetX[iColumn >> nBits] + iY);\
pT[iB] = pT[1] = pT[iR]; \
} \
} \
break
switch (pII->bdBitDepth)
{
case BD_8:
fixupThumb(U8, (pII->cBitsPerUnit >> 3));
break;
case BD_16:
case BD_16S:
case BD_16F:
fixupThumb(U16, (pII->cBitsPerUnit >> 3) / sizeof(U16));
break;
case BD_32:
case BD_32S:
case BD_32F:
fixupThumb(U32, (pII->cBitsPerUnit >> 3) / sizeof(float));
break;
case BD_5:
case BD_10:
case BD_565:
default:
break;
}
}
#endif
return ICERR_OK;
}
/*************************************************************************
Read variable length byte aligned integer
*************************************************************************/
static size_t GetVLWordEsc(BitIOInfo* pIO, Int *iEscape)
{
size_t s;
if (iEscape)
*iEscape = 0;
s = getBit32(pIO, 8);
if (s == 0xfd || s == 0xfe || s == 0xff) {
if (iEscape)
*iEscape = (Int) s;
s = 0;
}
else if (s < 0xfb) {
s = (s << 8) | getBit32(pIO, 8);
}
else {
s -= 0xfb;
if (s) {
s = getBit32(pIO, 16) << 16;
s = (s | getBit32(pIO, 16)) << 16;
s <<= 16;
}
s |= (getBit32(pIO, 16) << 16);
s |= getBit32(pIO, 16);
}
return s;
}
//================================================================
Int readIndexTable(CWMImageStrCodec * pSC)
{
BitIOInfo* pIO = pSC->pIOHeader;
readIS_L1(pSC, pIO);
if(pSC->cNumBitIO > 0){
size_t *pTable = pSC->pIndexTable;
U32 iEntry = (U32)pSC->cNumBitIO * (pSC->WMISCP.cNumOfSliceMinus1H + 1), i;
// read index table header [0x0001] - 2 bytes
if (getBit32(pIO, 16) != 1)
return ICERR_ERROR;
//iBits = getBit16(pIO, 5) + 1; // how many bits per entry
for(i = 0; i < iEntry; i ++){
readIS_L1(pSC, pIO);
pTable[i] = GetVLWordEsc(pIO, NULL); // escape handling is not important since the respective band is not accessed
}
}
pSC->cHeaderSize = GetVLWordEsc(pIO, NULL); // escape handling is not important
flushToByte(pIO);
pSC->cHeaderSize += getPosRead(pSC->pIOHeader); // get header length
return ICERR_OK;
}
Int StrIODecInit(CWMImageStrCodec* pSC)
{
if(allocateBitIOInfo(pSC) != ICERR_OK){
return ICERR_ERROR;
}
attachISRead(pSC->pIOHeader, pSC->WMISCP.pWStream, pSC);
readIndexTable(pSC);
if(pSC->WMISCP.bVerbose){
U32 i, j;
printf("\n%d horizontal tiles:\n", pSC->WMISCP.cNumOfSliceMinus1H + 1);
for(i = 0; i <= pSC->WMISCP.cNumOfSliceMinus1H; i ++){
printf(" offset of tile %d in MBs: %d\n", i, pSC->WMISCP.uiTileY[i]);
}
printf("\n%d vertical tiles:\n", pSC->WMISCP.cNumOfSliceMinus1V + 1);
for(i = 0; i <= pSC->WMISCP.cNumOfSliceMinus1V; i ++){
printf(" offset of tile %d in MBs: %d\n", i, pSC->WMISCP.uiTileX[i]);
}
if(pSC->WMISCP.bfBitstreamFormat == SPATIAL){
printf("\nSpatial order bitstream\n");
}
else{
printf("\nFrequency order bitstream\n");
}
if(!pSC->m_param.bIndexTable){
printf("\nstreaming mode, no index table.\n");
}
else if(pSC->WMISCP.bfBitstreamFormat == SPATIAL){
for(j = 0; j <= pSC->WMISCP.cNumOfSliceMinus1H; j ++){
for(i = 0; i <= pSC->WMISCP.cNumOfSliceMinus1V; i ++){
size_t * p = &pSC->pIndexTable[j * (pSC->WMISCP.cNumOfSliceMinus1V + 1) + i];
if(i + j != pSC->WMISCP.cNumOfSliceMinus1H + pSC->WMISCP.cNumOfSliceMinus1V){
printf("bitstream size for tile (%d, %d): %d.\n", j, i, (int) (p[1] - p[0]));
}
else{
printf("bitstream size for tile (%d, %d): unknown.\n", j, i);
}
}
}
}
else{
for(j = 0; j <= pSC->WMISCP.cNumOfSliceMinus1H; j ++){
for(i = 0; i <= pSC->WMISCP.cNumOfSliceMinus1V; i ++){
size_t * p = &pSC->pIndexTable[(j * (pSC->WMISCP.cNumOfSliceMinus1V + 1) + i) * 4];
if(i + j != pSC->WMISCP.cNumOfSliceMinus1H + pSC->WMISCP.cNumOfSliceMinus1V){
printf("bitstream size of (DC, LP, AC, FL) for tile (%d, %d): %d %d %d %d.\n", j, i,
(int) (p[1] - p[0]), (int) (p[2] - p[1]), (int) (p[3] - p[2]), (int) (p[4] - p[3]));
}
else{
printf("bitstream size of (DC, LP, AC, FL) for tile (%d, %d): %d %d %d unknown.\n", j, i,
(int) (p[1] - p[0]), (int) (p[2] - p[1]), (int) (p[3] - p[2]));
}
}
}
}
}
return 0;
}
Int StrIODecTerm(CWMImageStrCodec* pSC)
{
detachISRead(pSC, pSC->pIOHeader);
free(pSC->m_ppBitIO);
free(pSC->pIndexTable);
return 0;
}
Int initLookupTables(CWMImageStrCodec* pSC)
{
static const U8 cbChannels[BDB_MAX] = {1, 1, 2, 2, 2, 4, 4, 4, (U8) -1, (U8) -1, (U8) -1 };
CWMImageInfo * pII = &pSC->WMII;
size_t cStrideX, cStrideY;
size_t w, h, i, iFirst = 0;
Bool bReverse;
// lookup tables for rotation and flipping
if(pSC->m_Dparam->cThumbnailScale > 1) // thumbnail
w = pII->cThumbnailWidth, h = pII->cThumbnailHeight;
else
w = pII->cWidth, h = pII->cHeight;
w += (pSC->m_Dparam->cROILeftX + pSC->m_Dparam->cThumbnailScale - 1) / pSC->m_Dparam->cThumbnailScale;
h += (pSC->m_Dparam->cROITopY + pSC->m_Dparam->cThumbnailScale - 1) / pSC->m_Dparam->cThumbnailScale;
switch(pII->bdBitDepth){
case BD_16:
case BD_16S:
case BD_5:
case BD_565:
case BD_16F:
cStrideY = pSC->WMIBI.cbStride / 2;
break;
case BD_32:
case BD_32S:
case BD_32F:
case BD_10:
cStrideY = pSC->WMIBI.cbStride / 4;
break;
default: //BD_8, BD_1
cStrideY = pSC->WMIBI.cbStride;
break;
}
switch(pII->cfColorFormat){
case YUV_420:
cStrideX = 6;
w >>= 1, h >>= 1;
break;
case YUV_422:
cStrideX = 4;
w >>= 1;
break;
default:
cStrideX = (pII->cBitsPerUnit >> 3) / cbChannels[pII->bdBitDepth];
break;
}
if(pII->bdBitDepth == BD_1 || pII->bdBitDepth == BD_5 || pII->bdBitDepth == BD_10 || pII->bdBitDepth == BD_565)
cStrideX = 1;
if(pII->oOrientation > O_FLIPVH) // rotated !!
i =cStrideX, cStrideX = cStrideY, cStrideY = i;
pSC->m_Dparam->pOffsetX = (size_t *)malloc(w * sizeof(size_t));
if(pSC->m_Dparam->pOffsetX == NULL || w * sizeof(size_t) < w)
return ICERR_ERROR;
/*
consider a row in the source image. if it becomes a reversed row in the target, or a reversed (upside-down)column
in the target, we have to reverse the offsets. bReverse here tells us when this happened.
*/
bReverse = (pII->oOrientation == O_FLIPH || pII->oOrientation == O_FLIPVH ||
pII->oOrientation == O_RCW_FLIPV || pII->oOrientation == O_RCW_FLIPVH);
if(!pSC->m_Dparam->bDecodeFullFrame) // take care of region decode here!
iFirst = (pSC->m_Dparam->cROILeftX + pSC->m_Dparam->cThumbnailScale - 1) / pSC->m_Dparam->cThumbnailScale;
for(i = 0; i + iFirst < w; i ++){
pSC->m_Dparam->pOffsetX[i + iFirst] = pII->cLeadingPadding + (bReverse ? (pSC->m_Dparam->bDecodeFullFrame ? w :
(pSC->m_Dparam->cROIRightX - pSC->m_Dparam->cROILeftX + pSC->m_Dparam->cThumbnailScale) / pSC->m_Dparam->cThumbnailScale / ((pII->cfColorFormat == YUV_420 || pII->cfColorFormat == YUV_422) ? 2 : 1)) - 1 - i : i) * cStrideX;
}
pSC->m_Dparam->pOffsetY = (size_t *)malloc(h * sizeof(size_t));
if(pSC->m_Dparam->pOffsetY == NULL || h * sizeof(size_t) < h)
return ICERR_ERROR;
/*
consider a column in the source image. if it becomes an upside-down column in the target, or a reversed row
in the target, we have to reverse the offsets. bReverse here tells us when this happened.
*/
bReverse = (pII->oOrientation == O_FLIPV || pII->oOrientation == O_FLIPVH ||
pII->oOrientation == O_RCW || pII->oOrientation == O_RCW_FLIPV);
if(!pSC->m_Dparam->bDecodeFullFrame) // take care of region decode here!
iFirst = (pSC->m_Dparam->cROITopY + pSC->m_Dparam->cThumbnailScale - 1) / pSC->m_Dparam->cThumbnailScale;
for(i = 0; i + iFirst < h; i ++){
pSC->m_Dparam->pOffsetY[i + iFirst] = (bReverse ? (pSC->m_Dparam->bDecodeFullFrame ? h :
(pSC->m_Dparam->cROIBottomY - pSC->m_Dparam->cROITopY + pSC->m_Dparam->cThumbnailScale) / pSC->m_Dparam->cThumbnailScale / (pII->cfColorFormat == YUV_420 ? 2 : 1)) - 1 - i : i) * cStrideY;
}
return ICERR_OK;
}
Void setROI(CWMImageStrCodec* pSC)
{
CWMImageInfo * pWMII = &pSC->WMII;
CWMIStrCodecParam * pSCP = &pSC->WMISCP;
// inscribed image size
pWMII->cWidth -= pSC->m_param.cExtraPixelsLeft + pSC->m_param.cExtraPixelsRight;
pWMII->cHeight -= pSC->m_param.cExtraPixelsTop + pSC->m_param.cExtraPixelsBottom;
pSC->m_Dparam->bSkipFlexbits = (pSCP->sbSubband == SB_NO_FLEXBITS);
pSC->m_Dparam->bDecodeHP = (pSCP->sbSubband == SB_ALL || pSCP->sbSubband == SB_NO_FLEXBITS);
pSC->m_Dparam->bDecodeLP = (pSCP->sbSubband != SB_DC_ONLY);
pSC->m_Dparam->cThumbnailScale = 1;
while(pSC->m_Dparam->cThumbnailScale * pWMII->cThumbnailWidth < pWMII->cWidth)
pSC->m_Dparam->cThumbnailScale <<= 1;
if(pSC->WMISCP.bfBitstreamFormat == FREQUENCY){
if(pSC->m_Dparam->cThumbnailScale >= 4)
pSC->m_Dparam->bDecodeHP = FALSE; // no need to decode HP
if(pSC->m_Dparam->cThumbnailScale >= 16)
pSC->m_Dparam->bDecodeLP = FALSE; // only need to decode DC
}
// original image size
pWMII->cWidth += pSC->m_param.cExtraPixelsLeft + pSC->m_param.cExtraPixelsRight;
pWMII->cHeight += pSC->m_param.cExtraPixelsTop + pSC->m_param.cExtraPixelsBottom;
/** region decode stuff */
pSC->m_Dparam->cROILeftX = pWMII->cROILeftX * pSC->m_Dparam->cThumbnailScale + pSC->m_param.cExtraPixelsLeft;
pSC->m_Dparam->cROIRightX = pSC->m_Dparam->cROILeftX + pWMII->cROIWidth * pSC->m_Dparam->cThumbnailScale - 1;
pSC->m_Dparam->cROITopY = pWMII->cROITopY * pSC->m_Dparam->cThumbnailScale + pSC->m_param.cExtraPixelsTop;
pSC->m_Dparam->cROIBottomY = pSC->m_Dparam->cROITopY + pWMII->cROIHeight * pSC->m_Dparam->cThumbnailScale - 1;
if(pSC->m_Dparam->cROIRightX >= pWMII->cWidth)
pSC->m_Dparam->cROIRightX = pWMII->cWidth - 1;
if(pSC->m_Dparam->cROIBottomY >= pWMII->cHeight)
pSC->m_Dparam->cROIBottomY = pWMII->cHeight - 1;
pSC->m_Dparam->bDecodeFullFrame = (pSC->m_Dparam->cROILeftX + pSC->m_Dparam->cROITopY == 0 &&
((pSC->m_Dparam->cROIRightX + 15) / 16 >= (pWMII->cWidth + 14) / 16) && ((pSC->m_Dparam->cROIBottomY + 15) / 16 >= (pWMII->cHeight + 14) / 16));
pSC->m_Dparam->bDecodeFullWidth = (pSC->m_Dparam->cROILeftX == 0 && ((pSC->m_Dparam->cROIRightX + 15) / 16 >= (pWMII->cWidth + 14) / 16));
// inscribed image size
pWMII->cWidth -= pSC->m_param.cExtraPixelsLeft + pSC->m_param.cExtraPixelsRight;
pWMII->cHeight -= pSC->m_param.cExtraPixelsTop + pSC->m_param.cExtraPixelsBottom;
if(pSC->WMISCP.bfBitstreamFormat == FREQUENCY && pWMII->bSkipFlexbits == TRUE)
pSC->m_Dparam->bSkipFlexbits = TRUE;
pSC->cTileColumn = pSC->cTileRow = 0;
}
Int StrDecInit(CWMImageStrCodec* pSC)
{
// CWMImageInfo * pWMII = &pSC->WMII;
COLORFORMAT cfInt = pSC->m_param.cfColorFormat;
COLORFORMAT cfExt = pSC->WMII.cfColorFormat;
size_t i;
/** color transcoding with resolution change **/
pSC->m_bUVResolutionChange = ((cfExt != Y_ONLY) && ((cfInt == YUV_420 && cfExt != YUV_420) ||
(cfInt == YUV_422 && cfExt != YUV_422))) && !pSC->WMISCP.bYUVData;
if(pSC->m_bUVResolutionChange){
pSC->pResU = (PixelI *)malloc((cfExt == YUV_422 ? 128 : 256) * pSC->cmbWidth * sizeof(PixelI));
pSC->pResV = (PixelI *)malloc((cfExt == YUV_422 ? 128 : 256) * pSC->cmbWidth * sizeof(PixelI));
if(pSC->pResU == NULL || pSC->pResV == NULL || (cfExt == YUV_422 ? 128 : 256) * pSC->cmbWidth * sizeof(PixelI) < pSC->cmbWidth){
return ICERR_ERROR;
}
}
if(allocatePredInfo(pSC) != ICERR_OK){
return ICERR_ERROR;
}
if(allocateTileInfo(pSC) != ICERR_OK)
return ICERR_ERROR;
if((pSC->m_param.uQPMode & 1) == 0){ // DC frame uniform quantization
if(allocateQuantizer(pSC->pTile[0].pQuantizerDC, pSC->m_param.cNumChannels, 1) != ICERR_OK)
return ICERR_ERROR;
setUniformQuantizer(pSC, 0);
for(i = 0; i < pSC->m_param.cNumChannels; i ++)
pSC->pTile[0].pQuantizerDC[i]->iIndex = pSC->m_param.uiQPIndexDC[i];
formatQuantizer(pSC->pTile[0].pQuantizerDC, (pSC->m_param.uQPMode >> 3) & 3, pSC->m_param.cNumChannels, 0, TRUE, pSC->m_param.bScaledArith);
}
if(pSC->WMISCP.sbSubband != SB_DC_ONLY){
if((pSC->m_param.uQPMode & 2) == 0){ // LP frame uniform quantization
if(allocateQuantizer(pSC->pTile[0].pQuantizerLP, pSC->m_param.cNumChannels, 1) != ICERR_OK)
return ICERR_ERROR;
setUniformQuantizer(pSC, 1);
if((pSC->m_param.uQPMode & 0x200) == 0) // use DC quantizer
useDCQuantizer(pSC, 0);
else{
for(i = 0; i < pSC->m_param.cNumChannels; i ++)
pSC->pTile[0].pQuantizerLP[i]->iIndex = pSC->m_param.uiQPIndexLP[i];
formatQuantizer(pSC->pTile[0].pQuantizerLP, (pSC->m_param.uQPMode >> 5) & 3, pSC->m_param.cNumChannels, 0, TRUE, pSC->m_param.bScaledArith);
}
}
if(pSC->WMISCP.sbSubband != SB_NO_HIGHPASS){
if((pSC->m_param.uQPMode & 4) == 0){ // HP frame uniform quantization
if(allocateQuantizer(pSC->pTile[0].pQuantizerHP, pSC->m_param.cNumChannels, 1) != ICERR_OK)
return ICERR_ERROR;
setUniformQuantizer(pSC, 2);
if((pSC->m_param.uQPMode & 0x400) == 0) // use LP quantizer
useLPQuantizer(pSC, 1, 0);
else{
for(i = 0; i < pSC->m_param.cNumChannels; i ++)
pSC->pTile[0].pQuantizerHP[i]->iIndex = pSC->m_param.uiQPIndexHP[i];
formatQuantizer(pSC->pTile[0].pQuantizerHP, (pSC->m_param.uQPMode >> 7) & 3, pSC->m_param.cNumChannels, 0, FALSE, pSC->m_param.bScaledArith);
}
}
}
}
if(pSC->WMISCP.cNumOfSliceMinus1V >= MAX_TILES || AllocateCodingContextDec(pSC, pSC->WMISCP.cNumOfSliceMinus1V + 1) != ICERR_OK){
return ICERR_ERROR;
}
if (pSC->m_bSecondary) {
pSC->pIOHeader = pSC->m_pNextSC->pIOHeader;
pSC->m_ppBitIO = pSC->m_pNextSC->m_ppBitIO;
pSC->cNumBitIO = pSC->m_pNextSC->cNumBitIO;
pSC->cSB = pSC->m_pNextSC->cSB;
}
setBitIOPointers(pSC);
return ICERR_OK;
}
Int StrDecTerm(CWMImageStrCodec* pSC)
{
size_t j, jend = (pSC->m_pNextSC != NULL);
for (j = 0; j <= jend; j++) {
if(pSC->m_bUVResolutionChange){
if(pSC->pResU != NULL)
free(pSC->pResU);
if(pSC->pResV != NULL)
free(pSC->pResV);
}
freePredInfo(pSC);
freeTileInfo(pSC);
FreeCodingContextDec(pSC);
if (j == 0) {
StrIODecTerm(pSC);
// free lookup tables for rotation and flipping
if(pSC->m_Dparam->pOffsetX != NULL)
free(pSC->m_Dparam->pOffsetX);
if(pSC->m_Dparam->pOffsetY != NULL)
free(pSC->m_Dparam->pOffsetY);
}
pSC = pSC->m_pNextSC;
}
return 0;
}
/*************************************************************************
Read header of image plane
*************************************************************************/
Int ReadImagePlaneHeader(CWMImageInfo* pII, CWMIStrCodecParam *pSCP,
CCoreParameters *pSC, SimpleBitIO* pSB)
{
ERR err = WMP_errSuccess;
pSC->cfColorFormat = getBit32_SB(pSB, 3); // internal color format
FailIf((pSC->cfColorFormat < Y_ONLY || pSC->cfColorFormat > NCOMPONENT), WMP_errUnsupportedFormat);
pSCP->cfColorFormat = pSC->cfColorFormat; // this should be removed later
pSC->bScaledArith = getBit32_SB(pSB, 1); // lossless mode
// subbands
pSCP->sbSubband = getBit32_SB(pSB, 4);
// color parameters
switch (pSC->cfColorFormat) {
case Y_ONLY:
pSC->cNumChannels = 1;
break;
case YUV_420:
pSC->cNumChannels = 3;
getBit32_SB(pSB, 1);
pII->cChromaCenteringX = (U8) getBit32_SB(pSB, 3);
getBit32_SB(pSB, 1);
pII->cChromaCenteringY = (U8) getBit32_SB(pSB, 3);
break;
case YUV_422:
pSC->cNumChannels = 3;
getBit32_SB(pSB, 1);
pII->cChromaCenteringX = (U8) getBit32_SB(pSB, 3);
getBit32_SB(pSB, 4);
break;
case YUV_444:
pSC->cNumChannels = 3;
getBit32_SB(pSB, 4);
getBit32_SB(pSB, 4);
break;
case NCOMPONENT:
pSC->cNumChannels = (Int) getBit32_SB(pSB, 4) + 1;
getBit32_SB(pSB, 4);
break;
case CMYK:
pSC->cNumChannels = 4;
break;
default:
break;
}
// float and 32s additional parameters
switch (pII->bdBitDepth) {
case BD_16:
case BD_16S:
case BD_32:
case BD_32S:
pSCP->nLenMantissaOrShift = (U8) getBit32_SB(pSB, 8);
break;
case BD_32F:
pSCP->nLenMantissaOrShift = (U8) getBit32_SB(pSB, 8);//float conversion parameters
pSCP->nExpBias = (I8) getBit32_SB(pSB, 8);
break;
default:
break;
}
// quantization
pSC->uQPMode = 0;
if(getBit32_SB(pSB, 1) == 1) // DC uniform
pSC->uQPMode += (readQuantizerSB(pSC->uiQPIndexDC, pSB, pSC->cNumChannels) << 3);
else
pSC->uQPMode ++;
if(pSCP->sbSubband != SB_DC_ONLY){
if(getBit32_SB(pSB, 1) == 0){ // don't use DC QP
pSC->uQPMode += 0x200;
if(getBit32_SB(pSB, 1) == 1) // LP uniform
pSC->uQPMode += (readQuantizerSB(pSC->uiQPIndexLP, pSB, pSC->cNumChannels) << 5);
else
pSC->uQPMode += 2;
}
else
pSC->uQPMode += ((pSC->uQPMode & 1) << 1) + ((pSC->uQPMode & 0x18) << 2);
if(pSCP->sbSubband != SB_NO_HIGHPASS){
if(getBit32_SB(pSB, 1) == 0){ // don't use LP QP
pSC->uQPMode += 0x400;
if(getBit32_SB(pSB, 1) == 1) // HP uniform
pSC->uQPMode += (readQuantizerSB(pSC->uiQPIndexHP, pSB, pSC->cNumChannels) << 7);
else
pSC->uQPMode += 4;
}
else
pSC->uQPMode += ((pSC->uQPMode & 2) << 1) + ((pSC->uQPMode & 0x60) << 2);
}
}
if(pSCP->sbSubband == SB_DC_ONLY)
pSC->uQPMode |= 0x200;
else if(pSCP->sbSubband == SB_NO_HIGHPASS)
pSC->uQPMode |= 0x400;
FailIf((pSC->uQPMode & 0x600) == 0, WMP_errInvalidParameter); // frame level QPs must be specified independently!
flushToByte_SB(pSB); // remove this later
Cleanup:
return WMP_errSuccess == err ? ICERR_OK : ICERR_ERROR;
}
/*************************************************************************
Read header of image, and header of FIRST PLANE only
*************************************************************************/
Int ReadWMIHeader(
CWMImageInfo* pII,
CWMIStrCodecParam *pSCP,
CCoreParameters *pSC)
{
U32 i;
ERR err = WMP_errSuccess;
Bool bTilingPresent, bInscribed, bTileStretch, bAbbreviatedHeader;
struct WMPStream* pWS = pSCP->pWStream;
SimpleBitIO SB = {0};
SimpleBitIO* pSB = &SB;
U8 szMS[8] = {0};
U32 cbStream = 0;
// U32 bits = 0;
// Int HEADERSIZE = 0;
assert(pSC != NULL);
//================================
// 0
/** signature **/
Call(pWS->Read(pWS, szMS, sizeof(szMS)));
FailIf(szMS != (U8 *) strstr((char *) szMS, "WMPHOTO"), WMP_errUnsupportedFormat);
//================================
Call(attach_SB(pSB, pWS));
// 8
/** codec version and subversion **/
i = getBit32_SB(pSB, 4);
FailIf((i != CODEC_VERSION), WMP_errIncorrectCodecVersion);
pSC->cVersion = i;
i = getBit32_SB(pSB, 4); // subversion
FailIf((i != CODEC_SUBVERSION &&
i != CODEC_SUBVERSION_NEWSCALING_SOFT_TILES && i != CODEC_SUBVERSION_NEWSCALING_HARD_TILES),
WMP_errIncorrectCodecSubVersion);
pSC->cSubVersion = i;
pSC->bUseHardTileBoundaries = FALSE;
if (pSC->cSubVersion == CODEC_SUBVERSION_NEWSCALING_HARD_TILES)
pSC->bUseHardTileBoundaries = TRUE;
pSCP->bUseHardTileBoundaries = pSC->bUseHardTileBoundaries;
// 9 primary parameters
bTilingPresent = (Bool) getBit32_SB(pSB, 1); // tiling present
pSCP->bfBitstreamFormat = getBit32_SB(pSB, 1); // bitstream layout
pII->oOrientation = (ORIENTATION)getBit32_SB(pSB, 3); // presentation orientation
pSC->bIndexTable = getBit32_SB(pSB, 1);
i = getBit32_SB(pSB, 2); // overlap
FailIf((i == 3), WMP_errInvalidParameter);
pSCP->olOverlap = i;
// 11 some other parameters
bAbbreviatedHeader = (Bool) getBit32_SB(pSB, 1); // short words for size and tiles
pSCP->bdBitDepth = (BITDEPTH) getBit32_SB(pSB, 1); // long word
pSCP->bdBitDepth = BD_LONG; // remove when optimization is done
bInscribed = (Bool) getBit32_SB(pSB, 1); // windowing
pSC->bTrimFlexbitsFlag = (Bool) getBit32_SB(pSB, 1); // trim flexbits flag
bTileStretch = (Bool) getBit32_SB(pSB, 1); // tile stretching flag
pSC->bRBSwapped = (Bool) getBit32_SB(pSB, 1); // red-blue swap flag
getBit32_SB(pSB, 1); // padding / reserved bit
pSC->bAlphaChannel = (Bool) getBit32_SB(pSB, 1); // alpha channel present
// 10 - informational
pII->cfColorFormat = getBit32_SB(pSB, 4); // source color format
pII->bdBitDepth = getBit32_SB(pSB, 4); // source bit depth
if(BD_1alt == pII->bdBitDepth)
{
pII->bdBitDepth = BD_1;
pSCP->bBlackWhite = 1;
}
// 12 - Variable length fields
// size
pII->cWidth = getBit32_SB(pSB, bAbbreviatedHeader ? 16 : 32) + 1;
pII->cHeight = getBit32_SB(pSB, bAbbreviatedHeader ? 16 : 32) + 1;
pSC->cExtraPixelsTop = pSC->cExtraPixelsLeft = pSC->cExtraPixelsBottom = pSC->cExtraPixelsRight = 0;
if (bInscribed == FALSE && (pII->cWidth & 0xf) != 0)
pSC->cExtraPixelsRight = 0x10 - (pII->cWidth & 0xF);
if (bInscribed == FALSE && (pII->cHeight & 0xf) != 0)
pSC->cExtraPixelsBottom = 0x10 - (pII->cHeight & 0xF);
// tiling
pSCP->cNumOfSliceMinus1V = pSCP->cNumOfSliceMinus1H = 0;
if (bTilingPresent) {
pSCP->cNumOfSliceMinus1V = getBit32_SB(pSB, LOG_MAX_TILES); // # of vertical slices along X axis
pSCP->cNumOfSliceMinus1H = getBit32_SB(pSB, LOG_MAX_TILES); // # of horizontal slices along Y axis
}
FailIf((pSC->bIndexTable == FALSE) && (pSCP->bfBitstreamFormat == FREQUENCY || pSCP->cNumOfSliceMinus1V + pSCP->cNumOfSliceMinus1H > 0),
WMP_errUnsupportedFormat);
// tile sizes
pSCP->uiTileX[0] = pSCP->uiTileY[0] = 0;
for(i = 0; i < pSCP->cNumOfSliceMinus1V; i ++){ // width in MB of vertical slices, not needed for last slice!
pSCP->uiTileX[i + 1] = (U32) getBit32_SB(pSB, bAbbreviatedHeader ? 8 : 16) + pSCP->uiTileX[i];
}
for(i = 0; i < pSCP->cNumOfSliceMinus1H; i ++){ // width in MB of vertical slices, not needed for last slice!
pSCP->uiTileY[i + 1] = (U32) getBit32_SB(pSB, bAbbreviatedHeader ? 8 : 16) + pSCP->uiTileY[i];
}
if (bTileStretch) { // no handling of tile stretching enabled as of now
for (i = 0; i < (pSCP->cNumOfSliceMinus1V + 1) * (pSCP->cNumOfSliceMinus1H + 1); i++)
getBit32_SB(pSB, 8);
}
// window due to compressed domain processing
if (bInscribed) {
pSC->cExtraPixelsTop = (U8)getBit32_SB(pSB, 6);
pSC->cExtraPixelsLeft = (U8)getBit32_SB(pSB, 6);
pSC->cExtraPixelsBottom = (U8)getBit32_SB(pSB, 6);
pSC->cExtraPixelsRight = (U8)getBit32_SB(pSB, 6);
}
if(((pII->cWidth + pSC->cExtraPixelsLeft + pSC->cExtraPixelsRight) & 0xf) + ((pII->cHeight + pSC->cExtraPixelsTop + pSC->cExtraPixelsBottom) & 0xf) != 0){
FailIf((pII->cWidth & 0xf) + (pII->cHeight & 0xf) + pSC->cExtraPixelsLeft + pSC->cExtraPixelsTop != 0, WMP_errInvalidParameter);
FailIf(pII->cWidth <= pSC->cExtraPixelsRight || pII->cHeight <= pSC->cExtraPixelsBottom, WMP_errInvalidParameter);
pII->cWidth -= pSC->cExtraPixelsRight, pII->cHeight -= pSC->cExtraPixelsBottom;
}
flushToByte_SB(pSB); // redundant
// read header of first image plane
FailIf(ReadImagePlaneHeader(pII, pSCP, pSC, pSB), WMP_errUnsupportedFormat);
// maybe UNALIGNED!!!
//================================
detach_SB(pSB);
pSCP->cbStream = cbStream - getByteRead_SB(pSB);
pSCP->uAlphaMode = (pSC->bAlphaChannel ? pSCP->uAlphaMode : 0);
pSCP->cChannel = pSC->cNumChannels;
if((pII->bdBitDepth == BD_5 || pII->bdBitDepth == BD_10 || pII->bdBitDepth == BD_565) &&
(pSCP->cfColorFormat != YUV_444 && pSCP->cfColorFormat != YUV_422 && pSCP->cfColorFormat != YUV_420 && pSCP->cfColorFormat != Y_ONLY))
return ICERR_ERROR;
Cleanup:
return WMP_errSuccess == err ? ICERR_OK : ICERR_ERROR;
}
//----------------------------------------------------------------
// streaming api init/decode/term
EXTERN_C Int ImageStrDecGetInfo(
CWMImageInfo* pII,
CWMIStrCodecParam *pSCP)
{
ERR err = WMP_errSuccess;
size_t cMarker;
CCoreParameters aDummy;
// mark position of start of data
Call(pSCP->pWStream->GetPos(pSCP->pWStream, &cMarker));
Call(ReadWMIHeader(pII, pSCP, &aDummy));
// rewind to start of data
Call(pSCP->pWStream->SetPos(pSCP->pWStream, cMarker));
return ICERR_OK;
Cleanup:
return ICERR_ERROR;
}
EXTERN_C Int WMPhotoValidate(
CWMImageInfo * pII,
CWMIStrCodecParam * pSCP)
{
CWMImageInfo cII;
CWMIStrCodecParam cSCP = *pSCP;
size_t cScale = 1;
if(ImageStrDecGetInfo(&cII, pSCP) != ICERR_OK)
return ICERR_ERROR;
// copy over un-overwritable ImageInfo parameters
pII->bdBitDepth = cII.bdBitDepth;
pII->cWidth = cII.cWidth;
pII->cHeight = cII.cHeight;
if(pII->cWidth == 0 || pII->cHeight == 0)
return ICERR_ERROR;
// copy over overwritable CodecParam parameters
pSCP->bVerbose = cSCP.bVerbose;
pSCP->cbStream = cSCP.cbStream;
pSCP->pWStream = cSCP.pWStream;
if(pSCP->uAlphaMode > 1) // something + alpha
pSCP->uAlphaMode = cSCP.uAlphaMode; // something + alpha to alpha or something transcoding!
// validate color transcoding
if(pSCP->cfColorFormat == NCOMPONENT)
pII->cfColorFormat = NCOMPONENT;
if(pSCP->cfColorFormat == CMYK && pII->cfColorFormat != Y_ONLY && pII->cfColorFormat != CF_RGB)
pII->cfColorFormat = CMYK;
if(pSCP->cfColorFormat == YUV_422 && pII->cfColorFormat == YUV_420)
pII->cfColorFormat = YUV_422;
if(pSCP->cfColorFormat == YUV_444 && (pII->cfColorFormat == YUV_422 || pII->cfColorFormat == YUV_420))
pII->cfColorFormat = YUV_444;
if(cII.cfColorFormat == CF_RGB && pII->cfColorFormat != Y_ONLY &&
pII->cfColorFormat != NCOMPONENT) // no guarantee that number of channels will be >= 3
pII->cfColorFormat = cII.cfColorFormat;
if(cII.cfColorFormat == CF_RGBE)
pII->cfColorFormat = CF_RGBE;
// validate thumbnail parameters
if(pII->cThumbnailWidth == 0 || pII->cThumbnailWidth > pII->cWidth)
pII->cThumbnailWidth = pII->cWidth;
if(pII->cThumbnailHeight == 0 || pII->cThumbnailHeight > pII->cHeight)
pII->cThumbnailHeight = pII->cHeight;
if((pII->cWidth + pII->cThumbnailWidth - 1) / pII->cThumbnailWidth != (pII->cHeight + pII->cThumbnailHeight - 1) / pII->cThumbnailHeight) {
while((pII->cWidth + cScale - 1) / cScale > pII->cThumbnailWidth &&
(pII->cHeight + cScale - 1) / cScale > pII->cThumbnailHeight && (cScale << 1))
cScale <<= 1;
}
else {
cScale = (pII->cWidth + pII->cThumbnailWidth - 1) / pII->cThumbnailWidth;
if (cScale == 0)
cScale = 1;
}
pII->cThumbnailWidth = (pII->cWidth + cScale - 1) / cScale;
pII->cThumbnailHeight = (pII->cHeight + cScale - 1) / cScale;
// validate region decode parameters
if(pII->cROIHeight == 0 || pII->cROIWidth == 0){
pII->cROILeftX = pII->cROITopY = 0;
pII->cROIWidth = pII->cThumbnailWidth;
pII->cROIHeight = pII->cThumbnailHeight;
}
if(pII->cROILeftX >= pII->cThumbnailWidth)
pII->cROILeftX = 0;
if(pII->cROITopY >= pII->cThumbnailHeight)
pII->cROITopY = 0;
if(pII->cROILeftX + pII->cROIWidth > pII->cThumbnailWidth)
pII->cROIWidth = pII->cThumbnailWidth - pII->cROILeftX;
if(pII->cROITopY + pII->cROIHeight > pII->cThumbnailHeight)
pII->cROIHeight = pII->cThumbnailHeight - pII->cROITopY;
return ICERR_OK;
}
/*************************************************************************
Initialization of CWMImageStrCodec struct
*************************************************************************/
static Void InitializeStrDec(CWMImageStrCodec *pSC,
const CCoreParameters *pParams, const CWMImageStrCodec *pSCIn)
{
// copy core parameters
memcpy (&(pSC->m_param), pParams, sizeof (CCoreParameters));
pSC->cbStruct = sizeof(*pSC);
pSC->WMII = pSCIn->WMII;
pSC->WMISCP = pSCIn->WMISCP;
pSC->cRow = 0;
pSC->cColumn = 0;
pSC->cmbWidth = (pSC->WMII.cWidth + 15) / 16;
pSC->cmbHeight = (pSC->WMII.cHeight + 15) / 16;
pSC->Load = outputMBRow; // output decoding result (ICC, etc)
pSC->Transform = pParams->cSubVersion == CODEC_SUBVERSION ?
invTransformMacroblock : invTransformMacroblock_alteredOperators_hard;
pSC->TransformCenter = pSC->Transform;
pSC->ProcessTopLeft = processMacroblockDec;
pSC->ProcessTop = processMacroblockDec;
pSC->ProcessTopRight = processMacroblockDec;
pSC->ProcessLeft = processMacroblockDec;
pSC->ProcessCenter = processMacroblockDec;
pSC->ProcessRight = processMacroblockDec;
pSC->ProcessBottomLeft = processMacroblockDec;
pSC->ProcessBottom = processMacroblockDec;
pSC->ProcessBottomRight = processMacroblockDec;
pSC->m_pNextSC = NULL;
pSC->m_bSecondary = FALSE;
}
/*************************************************************************
ImageStrDecInit
*************************************************************************/
Int ImageStrDecInit(
CWMImageInfo* pII,
CWMIStrCodecParam *pSCP,
CTXSTRCODEC* pctxSC)
{
static size_t cbChannels[BD_MAX] = {2, 4};
ERR err = WMP_errSuccess;
size_t cbChannel = 0, cblkChroma = 0;
size_t cbMacBlockStride = 0, cbMacBlockChroma = 0, cMacBlock = 0;
CWMImageStrCodec SC = {0};
CWMImageStrCodec *pSC = NULL, *pNextSC = NULL;
char* pb = NULL;
size_t cb = 0, i;
Bool bLossyTranscoding = FALSE;
Bool bUseHardTileBoundaries = FALSE; //default is soft tile boundaries
Bool bLessThan64Bit = sizeof(void *) < 8;
*pctxSC = NULL;
if(WMPhotoValidate(pII, pSCP) != ICERR_OK)
return ICERR_ERROR;
if(pSCP->sbSubband == SB_ISOLATED) // can not do anything with isolated bitstream
return ICERR_ERROR;
//================================================
SC.WMISCP.pWStream = pSCP->pWStream;
if (ReadWMIHeader(&SC.WMII, &SC.WMISCP, &SC.m_param) != ICERR_OK) {
return ICERR_ERROR;
}
bUseHardTileBoundaries = SC.WMISCP.bUseHardTileBoundaries;
if(SC.WMII.cfColorFormat == CMYK && pII->cfColorFormat == CF_RGB)
bLossyTranscoding = TRUE;
if(pSCP->cfColorFormat != CMYK && (pII->cfColorFormat == CMYK))
return ICERR_ERROR;
//================================================
SC.WMISCP = *pSCP;
SC.WMII = *pII;
// original image size
SC.WMII.cWidth += SC.m_param.cExtraPixelsLeft + SC.m_param.cExtraPixelsRight;
SC.WMII.cHeight += SC.m_param.cExtraPixelsTop + SC.m_param.cExtraPixelsBottom;
pII->cROILeftX += SC.m_param.cExtraPixelsLeft;
pII->cROITopY += SC.m_param.cExtraPixelsTop;
//================================================
cbChannel = cbChannels[SC.WMISCP.bdBitDepth];
cblkChroma = cblkChromas[SC.m_param.cfColorFormat];
cbMacBlockStride = cbChannel * 16 * 16;
cbMacBlockChroma = cbChannel * 16 * cblkChroma;
cMacBlock = (SC.WMII.cWidth + 15) / 16;
//================================================
cb = sizeof(*pSC) + (128 - 1) + sizeof(CWMDecoderParameters);
cb += (PACKETLENGTH * 4 - 1) + (PACKETLENGTH * 2 ) + sizeof(*pSC->pIOHeader);
i = (cbMacBlockStride + cbMacBlockChroma * (SC.m_param.cNumChannels - 1)) * 2; // i <= 2^15
if (bLessThan64Bit && ((i * (cMacBlock >> 16)) & 0xffffc000)) {
/** potential overflow - 32 bit pointers insufficient to address cache **/
return ICERR_ERROR;
}
cb += i * cMacBlock;
pb = malloc(cb);
if(pb == NULL)
return WMP_errOutOfMemory;
memset(pb, 0, cb);
//================================================
pSC = (CWMImageStrCodec*)pb; pb += sizeof(*pSC);
if(pSC == NULL)
return ICERR_ERROR;
// Set up perf timers
PERFTIMER_ONLY(pSC->m_fMeasurePerf = pSCP->fMeasurePerf);
PERFTIMER_NEW(pSC->m_fMeasurePerf, &pSC->m_ptEndToEndPerf);
PERFTIMER_NEW(pSC->m_fMeasurePerf, &pSC->m_ptEncDecPerf);
PERFTIMER_START(pSC->m_fMeasurePerf, pSC->m_ptEndToEndPerf);
PERFTIMER_START(pSC->m_fMeasurePerf, pSC->m_ptEncDecPerf);
PERFTIMER_COPYSTARTTIME(pSC->m_fMeasurePerf, pSC->m_ptEncDecPerf, pSC->m_ptEndToEndPerf);
pSC->m_Dparam = (CWMDecoderParameters*)pb; pb += sizeof(CWMDecoderParameters);
pSC->cbChannel = cbChannel;
//pSC->cNumChannels = SC.WMISCP.cChannel;
pSC->bUseHardTileBoundaries = bUseHardTileBoundaries;
//================================================
InitializeStrDec(pSC, &SC.m_param, &SC);
//================================================
// 2 Macro Row buffers for each channel
pb = ALIGNUP(pb, 128);
for (i = 0; i < pSC->m_param.cNumChannels; i++) {
pSC->a0MBbuffer[i] = (PixelI*)pb; pb += cbMacBlockStride * pSC->cmbWidth;
pSC->a1MBbuffer[i] = (PixelI*)pb; pb += cbMacBlockStride * pSC->cmbWidth;
cbMacBlockStride = cbMacBlockChroma;
}
//================================================
// lay 2 aligned IO buffers just below pIO struct
pb = (char*)ALIGNUP(pb, PACKETLENGTH * 4) + PACKETLENGTH * 2;
pSC->pIOHeader = (BitIOInfo*)pb; pb += sizeof(*pSC->pIOHeader);
// if interleaved alpha is needed
if (pSC->m_param.bAlphaChannel) {
SimpleBitIO SB = {0};
cbMacBlockStride = cbChannel * 16 * 16;
// 1. allocate new pNextSC info
//================================================
cb = sizeof(*pNextSC) + (128 - 1) + cbMacBlockStride * cMacBlock * 2;
// if primary image is safe to allocate, alpha channel is certainly safe
pb = malloc(cb);
if(pb == NULL)
return WMP_errOutOfMemory;
memset(pb, 0, cb);
//================================================
pNextSC = (CWMImageStrCodec*)pb; pb += sizeof(*pNextSC);
// read plane header of second image plane
Call(attach_SB(&SB, pSCP->pWStream));
InitializeStrDec(pNextSC, &SC.m_param, &SC);
ReadImagePlaneHeader(&pNextSC->WMII, &pNextSC->WMISCP, &pNextSC->m_param, &SB);
detach_SB(&SB);
// 2. initialize pNextSC
if(pNextSC == NULL)
return ICERR_ERROR;
pNextSC->m_Dparam = pSC->m_Dparam;
pNextSC->cbChannel = cbChannel;
//================================================
// 3. initialize arrays
// InitializeStrDec(pNextSC, &SC.m_param, &SC);
pNextSC->m_param.cfColorFormat = Y_ONLY;
pNextSC->m_param.cNumChannels = 1;
pNextSC->m_param.bAlphaChannel = TRUE;
//================================================
// 2 Macro Row buffers for each channel
pb = ALIGNUP(pb, 128);
pNextSC->a0MBbuffer[0] = (PixelI*)pb; pb += cbMacBlockStride * pNextSC->cmbWidth;
pNextSC->a1MBbuffer[0] = (PixelI*)pb;
//================================================
pNextSC->pIOHeader = pSC->pIOHeader;
//================================================
// 4. link pSC->pNextSC = pNextSC
pNextSC->m_pNextSC = pSC;
pNextSC->m_bSecondary = TRUE;
}
else
pSC->WMISCP.uAlphaMode = 0;
//================================================
FailIf((StrIODecInit(pSC) != ICERR_OK), WMP_errOutOfMemory);
FailIf((StrDecInit(pSC) != ICERR_OK), WMP_errOutOfMemory);
if (pNextSC) {
// 5. StrEncInit
FailIf((StrDecInit(pNextSC) != ICERR_OK), WMP_errOutOfMemory);
}
pSC->m_pNextSC = pNextSC;
//================================================
*pII = pSC->WMII;
*pSCP = pSC->WMISCP;
*pctxSC = (CTXSTRCODEC)pSC;
// original image size
pII->cROILeftX += SC.m_param.cExtraPixelsLeft;
pII->cROITopY += SC.m_param.cExtraPixelsTop;
if(pSC->WMII.cPostProcStrength){
initPostProc(pSC->pPostProcInfo, pSC->cmbWidth, pSC->m_param.cNumChannels);
if (pSC->m_param.bAlphaChannel)
initPostProc(pNextSC->pPostProcInfo, pNextSC->cmbWidth, pNextSC->m_param.cNumChannels);
}
PERFTIMER_STOP(pSC->m_fMeasurePerf, pSC->m_ptEncDecPerf);
Cleanup:
return WMP_errSuccess == err ? ICERR_OK : ICERR_ERROR;
}
Int ImageStrDecDecode(
CTXSTRCODEC ctxSC,
const CWMImageBufferInfo* pBI
#ifdef REENTRANT_MODE
, size_t *pcDecodedLines
#endif
)
{
CWMImageStrCodec* pSC = (CWMImageStrCodec*)ctxSC;
CWMImageStrCodec* pNextSC = pSC->m_pNextSC;
size_t cMBRow, k;
ImageDataProc ProcessLeft, ProcessCenter, ProcessRight;
ImageDataProc Transform = NULL;
const size_t iChromaElements = (pSC->m_param.cfColorFormat == YUV_420) ? 8 * 8
: ((pSC->m_param.cfColorFormat == YUV_422) ? 8 * 16 : 16 * 16);
if (sizeof(*pSC) != pSC->cbStruct)
{
return ICERR_ERROR;
}
//================================
PERFTIMER_START(pSC->m_fMeasurePerf, pSC->m_ptEncDecPerf);
pSC->WMIBI = *pBI;
#ifdef REENTRANT_MODE
if (0 == pSC->WMIBI.uiFirstMBRow)
{
setROI(pSC);
if (pNextSC) {
pNextSC->WMIBI = pSC->WMIBI;
setROI(pNextSC);
}
}
#else
setROI(pSC);
if (pNextSC) {
pNextSC->WMIBI = pSC->WMIBI;
setROI(pNextSC);
}
#endif // REENTRANT_MODE
// optimization flags can be defined only after ROI is set!
#if defined(WMP_OPT_SSE2) || defined(WMP_OPT_CC_DEC) || defined(WMP_OPT_TRFM_DEC)
StrDecOpt(pSC);
#endif // OPT defined
cMBRow = (pSC->m_Dparam->bDecodeFullFrame ? pSC->cmbHeight : ((pSC->m_Dparam->cROIBottomY + 16) >> 4));
#ifdef REENTRANT_MODE
if (0 == pSC->WMIBI.uiFirstMBRow)
{
if(initLookupTables(pSC) != ICERR_OK)
return ICERR_ERROR;
if (pNextSC && initLookupTables(pNextSC) != ICERR_OK)
return ICERR_ERROR;
}
#else
if(initLookupTables(pSC) != ICERR_OK)
return ICERR_ERROR;
if (pNextSC && initLookupTables(pNextSC) != ICERR_OK)
return ICERR_ERROR;
#endif // REENTRANT_MODE
#ifndef REENTRANT_MODE
if(pSC->WMII.bdBitDepth == BD_1){
size_t i;
for(i = 0; i < pSC->WMIBI.cLine; i ++)
memset(pSC->WMIBI.pv, 0, pSC->WMIBI.cbStride);
}
#endif
//================================
// top row
#ifdef REENTRANT_MODE
#else
pSC->cRow = 0;
ProcessLeft = pSC->ProcessTopLeft;
ProcessCenter = pSC->ProcessTop;
ProcessRight = pSC->ProcessTopRight;
Transform = pSC->m_param.cSubVersion == CODEC_SUBVERSION ?
invTransformMacroblock : invTransformMacroblock_alteredOperators_hard;
#endif // REENTRANT_MODE
#ifdef REENTRANT_MODE
for (pSC->cRow = pSC->WMIBI.uiFirstMBRow; pSC->cRow <= pSC->WMIBI.uiLastMBRow; pSC->cRow++)
{
// const COLORFORMAT cfExt = (pSC->m_param.cfColorFormat == Y_ONLY ? Y_ONLY : pSC->WMII.cfColorFormat);
if (0 == pSC->cRow)
{
ProcessLeft = pSC->ProcessTopLeft;
ProcessCenter = pSC->ProcessTop;
ProcessRight = pSC->ProcessTopRight;
Transform = pSC->m_param.cSubVersion == CODEC_SUBVERSION ?
invTransformMacroblock : invTransformMacroblock_alteredOperators_hard;
}
else if (cMBRow == pSC->cRow)
{
//================================
// bottom row
ProcessLeft = pSC->ProcessBottomLeft;
ProcessCenter = pSC->ProcessBottom;
ProcessRight = pSC->ProcessBottomRight;
Transform = pSC->m_param.cSubVersion == CODEC_SUBVERSION ?
invTransformMacroblock : invTransformMacroblock_alteredOperators_hard;
}
else { // middle rows
ProcessLeft = pSC->ProcessLeft;
ProcessCenter = pSC->ProcessCenter;
ProcessRight = pSC->ProcessRight;
Transform = pSC->TransformCenter;
}
#else
//================================
// central rows
for(pSC->cRow = 0; pSC->cRow <= cMBRow; pSC->cRow++)
{
#endif // REENTRANT_MODE
pSC->cColumn = 0;
initMRPtr(pSC);
/** zero out the transform coefficients (pull this out to once per MB row) **/
memset(pSC->p1MBbuffer[0], 0, sizeof(PixelI) * 16 * 16 * pSC->cmbWidth);
for (k = 1; k < pSC->m_param.cNumChannels; k++) {
memset(pSC->p1MBbuffer[k], 0, sizeof(PixelI) * iChromaElements * pSC->cmbWidth);
}
if (pSC->m_pNextSC != NULL) { // alpha channel
memset(pSC->m_pNextSC->p1MBbuffer[0], 0, sizeof(PixelI) * 16 * 16 * pSC->m_pNextSC->cmbWidth);
}
if(ProcessLeft(pSC) != ICERR_OK)
return ICERR_ERROR;
advanceMRPtr(pSC);
pSC->Transform = Transform;
for (pSC->cColumn = 1; pSC->cColumn < pSC->cmbWidth; ++pSC->cColumn)
{
if(ProcessCenter(pSC) != ICERR_OK)
return ICERR_ERROR;
advanceMRPtr(pSC);
}
pSC->Transform = pSC->m_param.cSubVersion == CODEC_SUBVERSION ?
invTransformMacroblock : invTransformMacroblock_alteredOperators_hard;
if(ProcessRight(pSC) != ICERR_OK)
return ICERR_ERROR;
if (pSC->cRow) {
if(pSC->m_Dparam->cThumbnailScale < 2 && (pSC->m_Dparam->bDecodeFullFrame ||
((pSC->cRow * 16 > pSC->m_Dparam->cROITopY) && (pSC->cRow * 16 <= pSC->m_Dparam->cROIBottomY + 16))))
pSC->Load(pSC); // bypass CC for thumbnail decode
if(pSC->m_Dparam->cThumbnailScale >= 2) // decode thumbnail
decodeThumbnail(pSC);
}
advanceOneMBRow(pSC);
swapMRPtr(pSC);
#ifdef REENTRANT_MODE
*pcDecodedLines = pSC->WMIBI.cLinesDecoded;
#else
if (pSC->cRow == cMBRow - 1) {
//================================
// bottom row
ProcessLeft = pSC->ProcessBottomLeft;
ProcessCenter = pSC->ProcessBottom;
ProcessRight = pSC->ProcessBottomRight;
Transform = pSC->m_param.cSubVersion == CODEC_SUBVERSION ?
invTransformMacroblock : invTransformMacroblock_alteredOperators_hard;
}
else {
ProcessLeft = pSC->ProcessLeft;
ProcessCenter = pSC->ProcessCenter;
ProcessRight = pSC->ProcessRight;
Transform = pSC->TransformCenter;
}
#endif // REENTRANT_MODE
}
#ifndef REENTRANT_MODE
fixup_Y_ONLY_to_Others(pSC, pBI);
#endif // REENTRANT_MODE
PERFTIMER_STOP(pSC->m_fMeasurePerf, pSC->m_ptEncDecPerf);
return ICERR_OK;
}
Int ImageStrDecTerm(
CTXSTRCODEC ctxSC)
{
CWMImageStrCodec* pSC = (CWMImageStrCodec*)ctxSC;
if (NULL == pSC)
{
return ICERR_OK;
}
if (sizeof(*pSC) != pSC->cbStruct)
{
return ICERR_ERROR;
}
PERFTIMER_START(pSC->m_fMeasurePerf, pSC->m_ptEncDecPerf);
StrDecTerm(pSC);
PERFTIMER_STOP(pSC->m_fMeasurePerf, pSC->m_ptEncDecPerf);
PERFTIMER_REPORT(pSC->m_fMeasurePerf, pSC);
PERFTIMER_DELETE(pSC->m_fMeasurePerf, pSC->m_ptEncDecPerf);
PERFTIMER_DELETE(pSC->m_fMeasurePerf, pSC->m_ptEndToEndPerf);
free(pSC);
return ICERR_OK;
}