JVET-AI0230 AHG14: Fix for mismatch between encoding and decoding in the NNVC with palette on

a method is proposed to fix the mismatch problem between encoding and decoding reconstruction by saving the reconstruction samples as prediction samples for palette mode.

can you provide in comment a command line testing this MR (I am assuming you just activate palette?)

You are right. The command line I used is as follows.

enc_cmd="bin/EncoderAppStatic -c cfg/encoder_randomaccess_nnvc.cfg --InputFile=sequences/SlideEditing.yuv -c cfg/per-class/classF.cfg --BitstreamFile=SlideEditing.bin --ReconFile=SlideEditing_rec.yuv --SourceWidth=1280 --SourceHeight=720 --QP=22 --FrameSkip=0 --FramesToBeEncoded=300 --IntraPeriod=32 --FrameRate=30 --InputBitDepth=8 --PLT=1 --NnIntraPred=0 --NnlfOption=3 --TemporalSubsampleRatio=1 --Level=3.1 >> result_SlideEditing.txt 2>&1"

dec_cmd="bin/DecoderAppStatic -b SlideEditing.bin -o SlideEditing_decrec.yuv --NnlfModelName=models/nnlf_lop3_model_int16.sadl >> result_SlideEditing_dec.txt"

In addition, the sadl version I use is "sadl-8f2a0707f493cfc5aa8e605096a4dc5f681969fa"

resolved all threads

merged

mentioned in commit 4a6727a6