An algorithm of image retrieval based on content similarity

To improve the performance of image retrieval, we propose an image retrieval method based on similarity calculation of visual objects in images. Firstly, one image is segmented and clustered several visual objects. By means of the keyword library, we calculate one probability vector, which is composed by probability of every visual object of the image and every keyword of the keyword library. So a total probability vector of one image can been attain. Then we calculate a content similarity according to two images´ probability vectors, thereby we can realize image retrieval. The end, Ground Truth Database is adopted experimental image library in this paper. The method attains good effect based on experimental result.