基于小波变换的图像处理方法是目前的主流方法，而对图像特征的多尺度统计建模则是图像压缩、去噪、分割、纹理分析与合成等统计应用的关键问题。本文综述了图像的多尺度统计模型，包括边缘分布模型以及层内、层间和混合相关模型，分析了各模型的优缺点，给出了对各种相关模型捕捉系数间相关性能力的归一化量测。最后，简单介绍了基于多尺度几何分析的统计图像模型，并对多尺度统计建模的前景进行了展望。

The algorithms based on wavelet transform have been very popular in image processing applications such as image compression, denoising, segmentation, texture analysis and synthesis. Multiscale statistical models for image characteristic are the key problems for these applications. This paper reviewed the statistical models for images in wavelet domain. Firstly, the marginal models for non-Gaussian distribution of image wavelet coefficients were studied, then the dependency models including interscale, intrascale and composite dependencies were analyzed, and the paper indicated the advantages and disadvantages of the models and gave normalized measures for the abilities of different dependency models to capture the dependencies between coefficients. At last, image statistical models based on multiscale geometric analysis were introduced in brief, and the possible future work is pointed out.