Image retrieval based on optimal matching with block histogram

This paper presents a new algorithm for content based image retrieval using color characteristics. The algorithm takes the Euclidean distance of color histogram between two images as the image similarity and consists of the following steps. First, it segments image into several blocks with the same size, and calculates the Euclidean distance of the color histogram between two arbitrary image blocks respectively belonging to two different images. Next, a weighted complete bipartite graph is constructed. The value of similarity for two images equals to the total cost of minimum cost perfect matching for the constructed graph. Then, the retrieval algorithm outputs the image sequences ordered by the value of similarity. Finally, a novel ideal is developed based on the faster speed of the histogram algorithm to reduce the computational complexity. As a result, compared with other color histogram-based algorithms, it shows that the presented retrieval method has better retrieval efficiency and is robust to rotation, translation and local distortion.