一种基于颜色连通的图像纹理检索新方法

A Novel Image Retrieval Method Using Texture Features Based on Color-Connected Regions

提出并实现一种结合图像颜色连通区域信息及其纹理特征的图像检索新方法 .首先提取图像的分块主颜色 ,根据提出的相关颜色定义 ,搜索确定图像中的颜色连通区域集 .然后 ,提取图像中各颜色连通区域对应的四种颜色共生矩阵特征 ,利用针对该特征设计的图像相似性度量函数实现基于内容的图像检索 .实验结果表明 ,该方法能有效结合图像的纹理信息及其颜色构成和分布信息 ,具有良好的检索效果和性能 .

This paper proposes a novel image retrieval method integrating image region information and the corresponding texture features.Firstly,the method extracts the dominant color of each image block to search the color-connected regions in the image according to a new definition of correlated color.Afterwards,the texture features corresponding to each color-connected region are computed and represented as color co-occurrence matrices,for which a specific similarity measurement is designed to retrieve images based on their contents.Experimental results have shown that the proposed method has sound and robust retrieval performance by integrating color composition and color spatial information into image texture descriptors.