一种基于小波多分辨率特征的彩色纹理图像检索

Color Texture Image Retrieval Based on Wavelet Transform Multiresolution Features

通过分析图像YCbCr空间的纹理与色彩特征,对色彩图像进行M带小波分解,提出了一种基于小波多分辨率特征和关系函数来计算图像相似度的彩色纹理图像检索方法.将小波分解后的16个子带系数作为图像的原始特征;运用模糊C均值算法对原始特征聚类分析,最后采用关系函数对查询图像和目标图像的模糊特征进行相似度计算,得到用户需求的图像.实验表明,该方法可以实现有效的图像特征提取,并能很好的检索出用户需求的图像.

A wavelet based muhiresolution features and relative function combined algorithm to compute the similarity of images in color and texture image retrieval is proposed in this paper. RGB color space is transformed into YCbCr space to keep independent of texture and color features, and then the images decomposition concludes 16 sub-bands; Fuzzy C-means clustering algorithm is used to analysis sub-band images and calculating similarity between query image and databases through relative function, and get the images the users need. The results were compared to other methods used in retrieval system.