基于小波变换的启发式分形编码技术

According to the vision characteristics of human being, a heuristic fractal coding of wavelet domain is systematically studied in this project by using the features of wavelet coefficients. That is, the wavelet coefficient domain is heuristically partitioned by using its hierarchy and directionality. After defining similar classes for coefficient blocks, an effective heuristic search based on similar classes is proposed. Also, block matching is sped up using the energy distribution of wavelet coefficients. Therefore, both of wavelet transform and fractal coding are combined completely, and an effective and efficient image compression technology is brought forward. Furthermore, another image compression algorithm based on zerotree and bit plane is proposed, which combines zerotree, bit plane and arithmetic coding together. The project also present a multi-scale and multi-resolution optical flow computation algorithm based on analytic wavelet and M-estimator. A new fast algorithm for video motion estimation is proposed, which takes use of the equality of motion vectors between neighbor blocks and the directionality of SAD distribution.

根据人类的视觉特性，利用小波系数的特点，系统研究小波域的启发式分形编码技术，既通过发现小波系数的层次性和方向性，提出小波域的启发式分割方法；在定义系数块的相似类基础上，提出相似类的启发式块搜索；根据系数的能量分布特征，研究加速块匹配的方法。从而实现小波变换和分形编码的有机结合，研究出一种新的高效图象压缩技术。