Abstract: An anisotropic diffusion scheme as a preprocessing step for lossy image compression，based on wavelet transform，is presented.This scheme combines anisotropic diffusion superiority with human vision characterization of wavelet transform.It leads to locally coarse grids in areas of resulting smooth image intensity，while enhanced edges are still resolved on fine grid level.Experiments show that for high lossy compression this new diffusion preprocessing results in better quality of the decoded image.

Key words: anisotropic diffusion, wavelet transform, PDE, image compression

摘要： 结合小波变换，提出了图像有损压缩预处理的各向异性扩散方案。各向异性扩散在图像平滑中，其不但能够较好地抑制噪声，而且能够很好地保留图像原有的边缘和纹理特征，而小波变换又具有人类的视觉特性。该方案先对图像作小波变换，再根据各子图像的特点，用不同的各向异性扩散方法处理子图，最后压缩重构图像。实验表明，对有损压缩来讲，与已有的方法比较，该方案具有更好的解码质量和更高的压缩比。

关键词: 各向异性扩散, 小波变换, 偏微分方程, 图像压缩