HEVC/H265 HM10 0 分析TAppDecTop cpp HEVC

须知少年凌云志，曾许人间第一流。这篇文章主要讲述HEVC/H265 HM10 0 分析TAppDecTop cpp相关的知识，希望能为你提供帮助。
在TAppDecTop.cpp ,最重要的是decode 函数，下面将对其进行分析，是解码上层的一个重要函数。
代码如下，代码后将进行分析。

Void TAppDecTop::decode() { Intpoc; TComList< TComPic*> * pcListPic = NULL; ifstream bitstreamFile(m_pchBitstreamFile, ifstream::in | ifstream::binary); if (!bitstreamFile) { fprintf(stderr, " failed to open bitstream file `%s‘ for reading ", m_pchBitstreamFile); exit(EXIT_FAILURE); }InputByteStream bytestream(bitstreamFile); // create & initialize internal classes xCreateDecLib(); xInitDecLib(); m_iPOCLastDisplay += m_iSkipFrame; // set the last displayed POC correctly for skip forward.// main decoder loop Bool recon_opened = false; // reconstruction file not yet opened. (must be performed after SPS is seen)while (!!bitstreamFile) { /* location serves to work around a design fault in the decoder, whereby * the process of reading a new slice that is the first slice of a new frame * requires the TDecTop::decode() method to be called again with the same * nal unit. */ streampos location = bitstreamFile.tellg(); AnnexBStats stats = AnnexBStats(); Bool bPreviousPictureDecoded = false; vector< uint8_t> nalUnit; InputNALUnit nalu; byteStreamNALUnit(bytestream, nalUnit, stats); // call actual decoding function Bool bNewPicture = false; if (nalUnit.empty()) { /* this can happen if the following occur: *- empty input file *- two back-to-back start_code_prefixes *- start_code_prefix immediately followed by EOF */ fprintf(stderr, "Warning: Attempt to decode an empty NAL unit "); } else { read(nalu, nalUnit); if( (m_iMaxTemporalLayer > = 0 & & nalu.m_temporalId > m_iMaxTemporalLayer) || !isNaluWithinTargetDecLayerIdSet(& nalu)) { if(bPreviousPictureDecoded) { bNewPicture = true; bPreviousPictureDecoded = false; } else { bNewPicture = false; } } else { bNewPicture = m_cTDecTop.decode(nalu, m_iSkipFrame, m_iPOCLastDisplay); if (bNewPicture) { bitstreamFile.clear(); /* location points to the current nalunit payload[1] due to the * need for the annexB parser to read three extra bytes. * [1] except for the first NAL unit in the file *(but bNewPicture doesn‘t happen then) */ bitstreamFile.seekg(location-streamoff(3)); bytestream.reset(); } bPreviousPictureDecoded = true; } } if (bNewPicture || !bitstreamFile) {m_cTDecTop.executeLoopFilters(poc, pcListPic); printf(" poc =%d ",poc); }if( pcListPic ) { printf(" naluType =%d ",nalu.m_nalUnitType); if ( m_pchReconFile & & !recon_opened ) { if (!m_outputBitDepthY) { m_outputBitDepthY = g_bitDepthY; } if (!m_outputBitDepthC) { m_outputBitDepthC = g_bitDepthC; }m_cTVideoIOYuvReconFile.open( m_pchReconFile, true, m_outputBitDepthY, m_outputBitDepthC, g_bitDepthY, g_bitDepthC ); // write mode recon_opened = true; } if ( bNewPicture & & (nalu.m_nalUnitType == NAL_UNIT_CODED_SLICE_IDR || nalu.m_nalUnitType == NAL_UNIT_CODED_SLICE_IDR_N_LP || nalu.m_nalUnitType == NAL_UNIT_CODED_SLICE_BLA_N_LP || nalu.m_nalUnitType == NAL_UNIT_CODED_SLICE_BLANT || nalu.m_nalUnitType == NAL_UNIT_CODED_SLICE_BLA ) ) { xFlushOutput( pcListPic ); } // write reconstruction to file if(bNewPicture) { xWriteOutput( pcListPic, nalu.m_temporalId ); } } }xFlushOutput( pcListPic ); // delete buffers m_cTDecTop.deletePicBuffer(); // destroy internal classes xDestroyDecLib(); }

代码xCreateDecLib 和 xInitDecLib 重要是初始化四叉树和解码需要的全局变量和申请内存。
当提供一个码流文件的文件名后，进行ifstream bitstreamFile(m_pchBitstreamFile, ifstream::in | ifstream::binary); 以二进制方式打开文件名，码流以字节方式InputByteStream bytestream(bitstreamFile); 进行读操作。

我们都知道，HEVC/H265 是以NAL方式组织数据的，解析VPS,SPS,PPS,SEI,SEI_SUFFIX 后，其他的是一个个slice的NAL数据，而Deblocking & SAO Filters 等滤波是对整个picuture进行滤波操作，出现从第二帧开始，每帧的第一个slice两次进行解析，这是参考软件的一个bug或不好的地方，其实完全可以知道是否是最后一个slice，不必进行两次打开。
所以出现：
bitstreamFile.clear();
bitstreamFile.seekg(location-streamoff(3));
bytestream.reset();

大家有兴趣，可以先增加一个变量，判断是否是最后一个slice，就不需要执行上面代码了。
【HEVC/H265 HM10 0 分析TAppDecTop cpp】
如下代码是从来不会执行，因为HEVC/H265没有cavlc，也就不会有slice part A ，slice part B，slice part C ，是现实的编解码告诉了设计者，slice part A ，slice part B，slice part C 没有人使用，就被抛弃了，实际的编解码从来没有实现slice part A ，slice part B，slice part C 等编解码的。
if( (m_iMaxTemporalLayer > = 0 & & nalu.m_temporalId > m_iMaxTemporalLayer) || !isNaluWithinTargetDecLayerIdSet(& nalu) )
{
if(bPreviousPictureDecoded)
{
bNewPicture = true;
bPreviousPictureDecoded = false;
}
else
{
bNewPicture = false;
}
}

解码和滤波后当然就是输出重构数据，xWriteOutput( pcListPic, nalu.m_temporalId )，如果slice是NAL_UNIT_CODED_SLICE_IDR，NAL_UNIT_CODED_SLICE_IDR_N_LP，NAL_UNIT_CODED_SLICE_BLA_N_LP，NAL_UNIT_CODED_SLICE_BLANT，NAL_UNIT_CODED_SLICE_BLA中的一种，将解码产生的picture全部清空，没有任何参考帧，相当于一个新的sequence进行解码了。

其他的代码很简单，请自己分析。

再分享一下我老师大神的人工智能教程吧。零基础！通俗易懂！风趣幽默！还带黄段子！希望你也加入到我们人工智能的队伍中来！https://blog.csdn.net/jiangjunshow