基于非下采样Contourlet信息熵的纹理图像检索

Texture Image Retrieval Based on Nonsubsampled Contourlet and Information Entropy

提出了一种基于非下采样Contourlet信息熵的纹理图像检索新方法.首先对原始图像进行三层非下采样Contourlet分解;然后计算各个子带的能量,并对其进行排序、合并,得到纹理直方图;最后以纹理直方图为概率密度函数,并用改进的信息熵方法求得函数的熵,以此作为纹理特征进行检索.实验结果表明,该方法与当前基于多分辨率的纹理图像检索方法相比具有更高的查全率和查准率.

A new method for texture image retrieval based on nonsubsampled Contourlet and information entropy was proposed. Firstly, image decomposition was conducted with nonsubsampied Contourlet transform. Then the energy of each sub-image was computed, and the sub-image was merged through its energy in order to get the histogram. Finally the information entropy was computed with the histogram. The proposed image retrieval method＇s performance was better than that of the approach of the traditional method based on the multiresolution theory.