摘要
本文提出了一种基于小波分解的快速分形图像编码方法．先对子图像块进行金字塔Haar变换；然后从低频段开始依次累加误差并与阈值比较，大于阈值则结束匹配由于各频带的数据量和能量分布不均匀，许多搜索过程可以提前结束．文中只需用对比度比例因子来估计误差，仅对搜索成功的块才计算亮度平移值．此外，对四叉树剖分的块方法也作了进一步的改进实验表明在不影响图像质量的前提下，编码时间缩短了100倍以上．
A fast fractal image eding bed on wavelet decomthetion is presented. First the subimage blocks are decmposed into different channels by Pyramidal bar transform. Then from the lower channel, error is accumulated and compared to the threshold TOL step by step. Because the data and energy of an image are not uniform, many searching processes can be ended early. In our scheme, only 't'is ed, and 's'is computed only for matched blocks. Furthermore, improvement for quadtree partition is also discussed. Simulation shows that the eding time is saved more than 100 times when the same coding result is kept.
出处
《电子学报》
EI
CAS
CSCD
北大核心
1998年第8期37-42,共6页
Acta Electronica Sinica
基金
浙江省科委96年度科技资助
关键词
分形
图像编码
金字塔Haar变换
Fractal, Image eding, Wavelet, Pyramidal Haar transform
