分形视频压缩的一种快速算法

在扩展的四叉树视频处理的基础上，分别对环形预测映射和帧间非压缩映射方法进行了改进，并且设计了将它们相结合的视频序列分形编码新方法。每个值域块由以往相邻的域块改成由相隔的图像帧中的相同尺寸的定义域块来逼近，而中间块则根据时间轴上高度相关性用插值方法得到。从而在实际图像序列压缩编码中，能充分地开发出时间上的相关性，减少编码时间。计算机数值试验的结果表明，该方法在低比特率40～250kbyte／s下，效果比较满意。

On Achieving Low Bit Rate and Quite Good Image Quality with Improved CPM and Improved NCIM

Kim and Lee proposed combining CPM (Circular Prediction Mapping) with NCIM (Noncontractive Interframe Mapping) to achieve big fractal video compression and transmission at low bit rate [7] . But their achievement brings also big reduction of image quality. We propose improving both CPM and NCIM in order to achieve big fractal video compression and transmission at low bit rate at the cost of only a small reduction of image quality. As a preparation for improving both CPM and NCIM, we make necessary changes in quadtree video partition method proposed by Fisher et al in Ref.6. In the improved CPM, each range block is approximated by a domain block of the same size; this is different from Ref.7, which approximates each range block with a domain block of bigger size. Improved NCIM differs from NCIM used in Ref.7 in that only every other frame is a directly utilized key-frame with the missing frames utilized indirectly through interpolation of directly utilized key-frames. Computer simulation results show preliminarily that our improvements in both CPM and NCIM do achieve big fractal video compression and transmission at low bit rate in the range of 40-250 kbyte/s at the cost of only a small reduction in image quality.