在CABAC中,较大的MV和Level首先二进制化为指数—哥伦布码,然后由算术编码引擎作为等概率符号逐位编码.指数-哥伦布码由前导1部分和定长编码部分组成.针对前导1部分和定长编码部分分别提出一种多二进制位并行解码算法.实验结果表明该算法相比逐位解码可以提供多达2.5倍的加速.%MV and Levels with big magnitudes is converted to exp-golomb code and coded as eq-probability symbols in CABAC. The exp-golomb code comprises of leading-one part and fix-length part Multi-bins parallel decoding algorithms are proposed in this paper for the leading-one and fix-length eq-probability symbols separately. Experiments show that the parallel decoding algorithm could give up to 2. Sx accelerating in decoding when compared to bin-wise decoding.
展开▼