摘要
利用Contourlet变换具有良好的稀疏特性及其能准确捕获图像中边缘信息的特性,提出一种基于Contourlet变换和不变矩的图像检索方法.利用Contourlet变换分解得到高、低频子带,计算各高频子带的信息熵和低频子带的7个不变矩构成图像的特征向量,采用不同权值的欧氏距离作为图像的相似度进行检索.实验结果表明,该方法具有较高的查准率,能够对纹理图像进行很好的分类和检索.
Taking advantage of the characteristics of good sparsity and effective capturing of the smooth contours in natural images of Contourlet transform,we have devised a new method of image retrieval based on Contourlet transform and moment invariants.After decomposing image by Contourlet transform,we have constructed feature vectors by calculating the entropy of high-frequency sub-bands and seven moment invariants of low-frequency sub-bands.The texture similarities of images were computed by the Euclidean's distance with different weights.The experiments show that the method has high precision rate and a good result has been obtained in image classification and retrieval.
出处
《徐州工程学院学报(自然科学版)》
CAS
2012年第1期48-51,共4页
Journal of Xuzhou Institute of Technology(Natural Sciences Edition)
基金
徐州工程学院青年教师基金(XKY2007319)
江苏省教育厅科研资助项目(07KJD520201)