Research Of Color-Base Image Retrieval System

In the field of the Content-Base image retrieval (CBIR) , the color feature as visual features is the most widely used . one hand because the color is the main perceptual features of the human eye's recognition. on the other hand, compared to other low-level features, color feature is highly robustness. This will do more further discussion about the color-based image retrieval method.The use of the color characteristics of image retrieval is a key step which is the description and extraction of color features, this paper According to the relevant knowledge of bipartite graph matching presents a new approach. This method based on the ideological sub-block color histogram, the image for each sub-block as a vertex graph, the similarity between the sub-block as the weight, thus constructed a weighted bipartite graph, looking for the right sub-block in the image corresponds to get the optimal matchin of bipartite graph ,afterin obtaining the similarity between all the sub-block in the two image,and get the right value of optimal matching through the KM algorithms, and thus obtain the similarity between two images. This method, taking into account the relationship between the spatial distribution of colors, they also inherited the rotation, translation invariance of the color features . The experiment proved that the method obtaines better retrieval results .