Contourlet变换结合了不可分离的方向滤波组,具备小波所不能表达的多方向特性,能有效捕获自然图像的边缘轮廓信息。本文分析了图像ontourlet系数的统计特征,并利用广义高斯函数对各子带系数层进行建模。将此模型应用于基于VisTex的自建纹理图像库,采用矩匹配估计法,提取模型参数集,运用K2L距离计算图像间的相似度。对800幅纹理图像进行检索,本文方法比传统小波方法的平均检索查准率高出约2％到10％不等。实验结果表明,该方法改进了导向纹理的描述-Transform Contourlet inseparable combination of the direction of filtering group, have not expressed by wavelet multi-directional properties, can effectively capture the edge of natural image contour information. This paper analyzes the image ontourlet coefficient statistical characteristics, and take advantage of the generalized Gaussian function of the sub-band coefficient modeling layer. Based on this model was applied to the self-VisTex texture image database, using moment matching estimation method, extract the model parameters set, K2L use of distance-based similarity between images. Of 800 texture image retrieval, wavelet method than the traditional method of the average retrieval precision by about 2 to 10 range. Experimental results show that the method improves the texture-oriented descr iption of