基于聚类主颜色和边缘直方图的图像检索方法

A Method of Image Retrieval Based on Clustered Domain Colors and Edge Histogram

图像的主颜色被广泛应用于图像检索中,但主颜色不能反映图像信息的空间分布。边缘直方图是在灰度图像上提取的,丢弃了图像的颜色信息。提出了一种综合利用聚类主颜色和边缘信息进行图像检索的方法。首先利用K均值聚类算法得到图像的主颜色,然后利用大津算法分割图像,利用Sobel算子提取目标和背景之间的显著的边缘特征信息构造边缘直方图,最后综合利用聚类主颜色和边缘直方图进行检索。既可以利用主色调信息,又能利用边缘的特征信息来反映图像信息的空间分布。实验结果表明,该方法可以有效地提高检索精度。

Domain colors of images are widely applied to image retrieval, but domain colors can not reflect spatial distribution of images＇ information. Edge histogram is extracted from gray images and discards color information of images. A new image retrieval method which uses clustered domain colors and edge information synthetically is put forward. First, domain colors of images are got by Kmeans clustering algorithm. Second, images are segmented by Otsu algorithm and edge histogram is constructed by using marked edge characteristic information between target and background which is extracted by Sobel operator. Third, retrieval is carried out by using clustered domain colors and edge histogram synthetically. Not only information of domain colors is used, but also characteristic information which reflects spatial distribution of images＇ information is used. The results of experiments show that this method can improve the precision of retrieval efficiently.