关于语法元素tcoeff_level_prediction_flag

(1)首先根据文豪的文档H.264_G.8_Flowchart_v1.pdf中关于SVC编码过程的描述：

当tcoeff_level_prediction_flag

0: spatial domain inter-layer prediction

1: frequency domain inter-layer prediction

 

(2)再查最新264标准T-REC-H.264-201003-I!!PDF-E.pdf，有以下相关的内容：

seq_tcoeff_level_prediction_flag： specifies the presence of the syntax element adaptive_tcoeff_level_prediction_flag in the subset sequence parameter set.

 

adaptive_tcoeff_level_prediction_flag： specifies the presence of tcoeff_level_prediction_flag in slice headers that refer to the subset sequence parameter set. When adaptive_tcoeff_level_prediction_flag is not present, it shall be inferred to be equal to 0.

seq_tcoeff_level_prediction_flag 和adaptive_tcoeff_level_prediction_flag是出现在subset sequence parameter set中的语法元素，含义从字面已经比较清楚，关键看tcoeff_level_prediction_flag的意义，它是出现在slice header中的语法元素。标准也对其含义进行了比较详细的描述，但是不太好理解。

 

(3)看JSVM SoftwareManual.doc中

2.4部分：SVC to AVC Bit-stream Rewriter

The SVC to AVC bitstream rewriter converts an SVC bitstream to an AVC bitstream. For each access unit in the input bitstream, the AVC bitstream rewriter converts all layer representations (in increasing order of DQId) that have either no_inter_layer_pred_flag equal to 1 or tcoeff_prediction_flag equal to 1 to an AVC picture. All other layer representations are ignored

(4)根据JSVM9.19代码

if( pcMbDataAccessBase && rcMbDataAccess.getMbData().getMbCbp() == 0

      && ( rcMbDataAccess.getSH().getSCoeffResidualPredFlag() || rcMbDataAccess.getSH().getTCoeffLevelPredictionFlag() ) // SpatialResolutionChangeFlag == 0

      && ( rcMbDataAccess.getMbData().getMbMode() == INTRA_BL || rcMbDataAccess.getMbData().getResidualPredFlag() ) )

 

从里面的注释// SpatialResolutionChangeFlag == 0可以比较清楚的推断，当seq_tcoeff_level_prediction_flag或tcoeff_level_prediction_flag为1的时候，即在变换域系数有预测的时候，应该是发生在没有空间分辨率变化的时候，即MGS和CGS层间预测的时候，

空间层间预测的时候，该值应该为0的。

 

 

待续。。。