基于颜色和空间特征的图像检索

Image Retrieval Based on Color-Spatial Feature

提出一种新的基于颜色和空间特征的图像检索算法.首先,将检索图像转换为HSV颜色空间并进行量化,提取环形颜色空间信息熵作为颜色空间分布特征.其次,计算每个像素点的多邻域量化颜色值的一、二阶中心矩,利用各阶统计矩的信息熵来表征图像颜色的局部空间特征.最后,对特征向量进行高斯归一化,采用特征向量的L1-norm距离计算彩色图像的相似度并进行图像检索.结果表明,该方法比CDE和Geostat算法具有较好的检索效果.

A new kind of color image retrieval algorithm based on color and spatial features is presented. At first, the color image is changed into quantized HSV color model, and the spatial-color information entropy can be obtained based on the annular color histogram. Secondly, the entropies of neighbor statistic moment are calculated. The spatial-color moments and the entropies of neighbor statistic moment are used as the character vector of color images. The Gaussian model is used to normalize the different sub-characters distance to the character vector. The similarity of the querying image and other images are computed by the L1-norm distance. Experiments indicate that this method has better performance than CDE and geostat.