Image Retrieval Technology Integration Of Color And Shape Feature

With the rapid development of IT, people are faced to more and more image information,in recent years. How to provide an effective way to fast and accurate query with the richconnotation of the image information has become a hot research field nowadays.Content-based image retrieval came into being to resolve it. In this paper, focusing on the wayto extracting and describe the feature of color and shape in CBIR. The main job is below:Analysis and compare with different color description models such as color histogram,color moment, color entropy, and set the color feature based on the L~*a~*b~*color space.First transformed the color space of an image from the RGB to the XYZ, then to L~*a~*b~*color space. Calculated the three low moment value of L~*and a~*and b~*to represent theimage’s color characteristics. L~*a~*b~*space has a wider color gamut performance rangethan RGB color space which is a device-independent color space. The color moments is moresimple and less dimension so it can greatly improve the computational efficiency of retrieval.Studying of the image segmented ways in object extraction, included the edge detectionand threshold segmentation method. Contrast to the characterization and region-based shapefeature description method based on the shape of the boundary. Obvious characteristics of theimage area in the image library for this article and using a single feature of the backgroundcolor descriptors to describe the shape feature based on Hu invariant moments. First iterativethresholding method to extract the target contour shape characteristics of the target area, andthen using the closing operation in morphological processing to external and internal filter,split, and then use the Hu invariant moments to describe the area. Hu invariant moments hasthe invariance with scale changes of translation, rotation, scaling. The amplitude of thesemoments reflect accurately reflect the shape of the object, and therefore this approach canaccurately describe the shape characteristics of the image.Retrieve images using color features or shape features alone are not comprehensive andlimitations. This article combines these two characteristics of integrated features to retrieve image. The weighted value between the them represent the color features and shape featuresin the image similarity retrieval. The experimental results show that the retrieval of integratedfeatures than the separate use of a particular feature to retrieve better. Poor precision andrecall rate of retrieval experiments for different similarity measure function, experimentalresults show that the weighted Euclidean distance is better.