摘要
本文通过对标准图像内在相关性的统计分析得出了自然图像具有方向自仿射性的结论,并首次提出了小波方向子树量化的概念。对分别属于水平、垂直和对角方向的尺度压缩因子是相互独立的情况采用小波方向子树量化进行了编/解码分析,同时深入研究了零树小波量化与小波方向子树量化之间的相互关系。实验证明,采用小波方向子树量化与零树量化的混合编码算法可使编码质量有较大的改善。
This paper first introduces the idea about orientation wavelet subtree quantization and constructs a hybrid image coding using orientation wavelet subtree fractal quantization and zerotree wavelet quantization. With this coding method the 'horizontal', 'vertical' and 'diagonal' scaling factors of quadtrees are treated independently according to the statistic analysis of test image orientation property and a result of a marked improvement in accuracy with respect to quantization for image compression is obtained. The coding results of this hybrid coding scheme outperform conventional fractal coders and Davis' SQS method.
关键词
小波方向子树
零树量化
图像编码
分形量化
Fractal wavelet transform, Orientation wavelet-subtree fractal quantization, Zerotree wavelet quantization, Hybrid image coding method
