一种基于颜色块直方图的图像检索方法

A New Approach to Image Retrieval Based on Histogram Combined with Color Block

提出了一种基于颜色直方图与颜色块相结合的图像检索的新方法。定义了一种新的四维图像特征矢量,它是由颜色直方图,同一种颜色的颜色块个数,同一种颜色块之间的平均距离和同一种颜色块内的平均距离组成。颜色直方图用来获取图像颜色分布的全局信息,利用后三个特征值来获取图像颜色的空间分布信息。通过对图像数据库的检索表明该算法明显优于直方图法。

In this paper, an approach to color-based image retrieval based on histogram combined with color block is proposed. A new 4 dimension vector of image, which is constituted by color histogram, the same color block number, the mean distance among the same color blocks and the mean distance inner color block, is defined. Color histogram, which can capture the global distribution of different intensities or color, is simply to calculate and especially it is not sensitive to image transport and rotation. The last three components are used to capture the spatial information of the image. Experimental results are given to show its ability of identifying images and retrieving similar images. Its performance is better than that of the histogram distance method.