prrt
9th January 2013, 21:55
Hi,
I develop H264 CABAC encoder. In CABAC decoder initialization process variable codIOffset is set equal to the value returned from read_bits( 9 ) interpreted as a 9 bit binary representation of an unsigned integer. But how to define these 9 bits in CABAC encoder?
For example, I encode MB=0 (first macroblock in slice) with mb_type = 3 (I_16x16_2_0_0) and intra_chroma_pred_mode=0. Bin string for mb_type will be '100010'. After context modelling and binary arithmetic encoding I will get bits '1111110' (it is result of PutBit()). After binarization and binary arithmetic coding of intra_pred_mode I will add to bitstream bits '10' so the resulted bitstream will be '111111010' (without any headers).
But when CABAC decoder will decode this bits it will set codIOffset = read_bits(9) = '111111010' = 506 and for decoding value of mb_type it will read next bits from bitstream. So in my encoder I must put first 9 bits. But how to calculate them?
I read H264 standard several times but I can't find information about it. Also I traced JM encoder and found that these 9 bits constantly updated during encoding the macroblock data, even when coding coeffs. But I can't understand on what algorithm.
Thanks in advance.
I develop H264 CABAC encoder. In CABAC decoder initialization process variable codIOffset is set equal to the value returned from read_bits( 9 ) interpreted as a 9 bit binary representation of an unsigned integer. But how to define these 9 bits in CABAC encoder?
For example, I encode MB=0 (first macroblock in slice) with mb_type = 3 (I_16x16_2_0_0) and intra_chroma_pred_mode=0. Bin string for mb_type will be '100010'. After context modelling and binary arithmetic encoding I will get bits '1111110' (it is result of PutBit()). After binarization and binary arithmetic coding of intra_pred_mode I will add to bitstream bits '10' so the resulted bitstream will be '111111010' (without any headers).
But when CABAC decoder will decode this bits it will set codIOffset = read_bits(9) = '111111010' = 506 and for decoding value of mb_type it will read next bits from bitstream. So in my encoder I must put first 9 bits. But how to calculate them?
I read H264 standard several times but I can't find information about it. Also I traced JM encoder and found that these 9 bits constantly updated during encoding the macroblock data, even when coding coeffs. But I can't understand on what algorithm.
Thanks in advance.