拼凑块矢量量化方法  

Puzzled block vector quantization algorithm

作　　者：孙文军[1] 孟中[2] 赵海鹰[3] 窦晓鸣[3] 

机构地区：[1]哈尔滨师范大学理化学院,黑龙江哈尔滨150025 [2]中国科学院长春光机所航测部,吉林长春130022 [3]上海交通大学光学工程研究所,上海200240

出　　处：《光电工程》2005年第11期63-67,共5页Opto-Electronic Engineering

基　　金：上海市科委光科项目(036105031)

摘　　要：提出了拼凑块矢量量化编码方法,利用四步部分匹配预测技术获得一个拼凑块,作为当前图像块的预测被编码,只需一位便可以完成当前图像块的编码。在此编码方案当中,根据预先定义的匹配失真阈值,图像可以有选择地利用拼凑块、动态码本及码本库编码。四步部分匹配预测技术利用优化了每个图像编码块的像素空间连续性,改善编码性能,降低了误差传递效率。实验结果表明PBVQ在0.27bpp时得到了30.92dB的峰值信噪比,明显优于IFMFSVQ的29.82dB,CSMVQ的28.56dB,GSMVQ的28.64dB,很大程度地改善了重构图像质量,尤其在低比特率下。A vector quantization algorithm based on puzzled block is presented, called as PBVQ （Puzzled Block Vector Quantization）. The PBVQ adopts four-step side-match prediction to offer a block called puzzled block as the prediction of the current image block to be coded, and only one bit is needed to encode the current image block. In this scheme, an input image block can be selectively encoded by the proposed puzzled block, the dynamic codebook or the super-codebook according to the predefined distortion threshold. The proposed algorithm uses a four-step side-match prediction technique to optimize the spatial continuity of the encoded block to improve the coding performance and reduce error propagation efficiency. Experimental results show that the PBVQ achieve 30.92dB at 0.27bpp which is better than 29.82dB of IFMFSVQ, 28.56dB of CSMVQ, 28.64dB of GSMVQ. Therefore, the proposed algorithm significantly improves image quality especially at low bit rate （about 0.25bpp）.

关 键 词：矢量量化 拼凑块 空间连续性 图像编码 

分 类 号：TN911.73[电子电信—通信与信息系统]