The Target Area-based Image Retrieval Techniques And Realization

The number of images is rapidly increasing with the development of internet. It has been a hot research to retrieve the needed images from lots of images accurately and effectively. The appearance of Content-Based Image Retrieval (CBIR) makes up the shortcomings of Text-Based Image Retrieval (TBIR). It bases on various content characteristics and focuses on the similarity-matching of these characteristic descriptions.Most of recent researches emphasize particularly on the global features, and neglect the differences between objects and background. In order to overcome the deficiencies of global features, some researches propose an image retrieval method based on object regions, which .uses the features of object regions to retrieve the resemble images.This paper describes an improved method about the object-region-based image retrieval. First, the method of boundary detection is used to get the edge information from grey image, which is transformed from color one. The scheme applies knowledge of morphology to enhance the edge image. Then the object regions are extracted from the binary edge image according to the connection of edge. Finally, the scheme uses unequally divided color to optimize the object regions. During the image retrieval, the features of color and shape from object regions are extracted and used in the image similarity matching process. The scheme uses integrated region matching method to get the retrieval result.Based on the improved method, this paper develops a system of object-region based image retrieval and compares the retrieval results with SIMPLIcity. The result of experiments shows that the improved method can abstract object regions efficiently and increase the precision.