Abstract: In video coding, DCT can visibly decrease the nature image’s relativity, but it is not good for motion residual. To improve the residuals coding efficiency, a new algorithm is proposed, that all non-zero quantized DCT coefficients of residual are divided into two classes (I and P), and they are classified for motion compensations. In MPEG4s coding quality experiments, the proposed method outperforms the standard MPEG4s by 2 dB upon. So the method can effectively increase coding efficiency for motion residual.

Key words: Discrete Cosine Transform(DCT), residual, motion compensation

摘要： 视频编码中离散余弦变换(DCT)对自然图像有很好的去相关性，而对运动残差的去相关性较差。针对DCT对残差编码效率较低的缺陷，该文提出一种基于残差分布特性的运动补偿算法，即将残差中所有非零量化DCT系数分为I/P系数，针对不同类型的DCT系数进行分类(帧内或帧间)运动补偿。在MPEG4的编码质量对比实验中，推荐算法比标准算法提高了2 dB以上，从而表明推荐算法能有效地提高运动残差的编码效率。

关键词: 离散余弦变换, 残差, 运动补偿

CLC Number: 

• TP393