综合分块颜色和基元共生矩阵的图像检索方法

Color Image Retrieval Based on the Combination of Block-color and Texture Features

提出了一种基于分块颜色直方图和纹理特征相结合的彩色图像检索方法。首先,该方法根据图像的背景和目标将其划分为两个区域,并对各个区域分别进行量化,提取颜色特征,计算图像之间的颜色距离;然后,用灰度共生矩阵提取图像的纹理特征并根据欧式距离来度量图像间的相似性。最后,综合利用颜色和纹理特征的加权来检索图像。经过对多种类别图像进行实验,检索结果都表明,与传统的检索方法相比,此方法能够很好的满足用户的需求,具有较高的查准率和查全率。

A new method for color image retrieval based on blocked color histogram and texture feature is presented in this paper. Firstly, the image is divided into two sub-blocks according to the differences between the object and background, and quantize and extract color features of each sub-block, compute the distance. Then the texture feature is proposed based on gray level co-occurrence matrix. Finally two features are combined to obtain the retrieval results. The results show that the proposed image retrieval performance is better than that of the traditional approaches.