基于幅相分离立体匹配的小波立体图像压缩

Wavelet Based Stereo Image Compression:Stereo Matching Using Amplitude and Phase Separately

针对立体图像压缩中视差估计问题,提出了幅度和相位分离的立体匹配方法.在压缩过程中,采用CONCOD立体图像编解码器结构,分别对左右图像进行小波多分辨率分解,并用幅相分离的立体匹配方法对左右图进行视差估计.根据各种补偿残差的大小,自适应调整预测块的大小和编码方案———帧间视差补偿、帧内预测补偿和帧内自编码.仿真实验证明了该算法在保持较高立体图像质量的同时可获得较高压缩比.

A new stereo matching algorithm using image amplitude and phase separately is proposed in order to realize disparity estimation in stereo image compression. In the procedure of stereo image compression, left and right images are decomposed based on multiscale wavelet, and match each other using the stereo matching algorithm with conditional coder structure. According to the compensated differences, prediction block size is adaptively changed, and coding scheme is also adaptively chosen among interframe disparity compensation, intraframe prediction compensation and intraframe coding. Simulation experiment proves that the algorithm can reach high compression ratio and keep quite good stereo image quality.