摘要： 针对目前分形小波混合图像压缩方法存在的不足，提出了一种新的分形小波混合图像压缩方法。同方向、不同分辨率的相邻子带图像之间进行分形搜索时，根据其系数的重要程度不同，采用不同的误差阈值与搜索步长，高分辨率子带的搜索步长大，误差阈值小，而低分辨率的子带则采用小步长和大误差阈值进行搜索。实验表明，相对于经典的基于小波树的分形图像压缩方法，该改进算法在不影响信噪比和解码图像质量的前提下，提高了压缩比和编码速度。

关键词: 小波, 分形, 图像压缩, 误差阈值

Abstract: By studying the limitations of existing hybrid fractal-wavelet image compression methods, this paper proposes a novel hybrid fractal-wavelet image compression algorithm. Briefly, it adopts diverse error thresholds as well as searching steps, considering varied significance pertaining to coefficients of adjacent subband images with the same orientation but different frequencies, when fractal searching is made among them. Great searching steps along with small error thresholds applies to high frequency subbands, while the contrary is suitable for low ones. Theoretical analysis and experiment results indicate that this algorithm largely increases the compression rate and the speed of encoding without reducing SNR and quality of decoded images, which is superior to the traditional wavelet tree based algorithms of fractal image compression.

Key words: wavelet, fractal, image compression, error thresholds

中图分类号: 

• TN 919.81