视频图像编码中的重排DCT方法  

Reorder discrete cosine transform method for video compression

作　　者：张贻雄[1] 陈凌宇[1] 王维东[2] 

机构地区：[1]厦门大学通信工程系,福建厦门361005 [2]浙江大学信息与电子工程学系,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2010年第9期1705-1710,共6页Journal of Zhejiang University：Engineering Science

基　　金：福建省科技厅重大专项资助项目(2007HZ0003)

摘　　要：为了提高离散余弦变换(DCT)在视频图像编码中的效率,根据DCT对非水平/非垂直残差信号较敏感的特点,从改变残差信号分布的角度出发,提出一种重排DCT方法.对待编码的残差图像块进行像素位置重排,使得残差信号的方向与水平或垂直方向相一致;对重排后的图像块进行2维DCT,从而减少DCT系数的高频分量,提高压缩效率.实验结果表明:在不考虑重排信息的情况下,重排DCT比传统DCT性能提高0.20～0.30dB;在考虑重排信息的情况下,重排DCT比传统DCT性能提高0.05～0.20dB.In traditional video compression, discrete cosine transform （DCT） is sensitive to the prediction error with non-horizontal or non-vertical orientations. To improve the performance of DCT in video compression, this work proposed a reorder DCT method, from the view point of modifying signal distribution. In the algorithm, reorder operation was first used to change the pixel position of the blocks to be transformed, and then 2D-DCT was applied. Experimental results show that the reorder DCT can reduce the number of DCT coefficients at high frequency. Compared to the traditional DCT, the reorder DCT achieved 0.20-0.30 dB peak signal-to-noise ratio （PSNR） gain without reorder information. With reorder information, 0.05-0.20 dB PSNR gain could be obtained.

关 键 词：视频图像编码 重排离散余弦变换 像素重排 

分 类 号：TN919.81[电子电信—通信与信息系统]