Based On Color And The Shape Characteristic Image Retrieval

With the development of multimedia technology and the rapid increase of digital libraries, image database, content-based image retrieval is currently a hot topic in the industry at home and abroad. CBIR is done according to the image color, shape, texture and other features or combinations of these features to query image, which is an effective integration of computer image processing and database technology. It not only makes full use of the information contained within the image, but also combines traditional database technology, so it is a very promising new technology in the theoretical study and practical applications. Shape is an important feature of the object and is the characteristics to describe the object’s structure, shape and contour. It is also one of the image characteristics which are difficult to describe. Because the automatic acquisition of object’s shape is related to the segmentation of interested targets in image, the shape-based image retrieval is a difficult research. Shape-based retrieval is generally applicable to the objects which are more easily identified. In this paper, size, shape moment and other local or global characteristics are used to an image.Based on the comprehensive analysis of the current content-based image retrieval method, this paper represents the extraction algorithms with the combination of shapes and colors, and achieves the appropriate retrieval system. What this paper studies are as follows:(1) First, review the situation, development trends and application status of CBIR technology at home and abroad, and compare the advantages and disadvantages of each system. Then introduce the shape feature extraction, in-depth studies the key of image retrieval. Also analyze the research status and limitations of the shape-based image retrieval technology.(2) According to the HSV color space, it achieves the quantitative description by separation means of hue H, saturation S and luminance V color. Using histogram to find the color feature, then find similar images according to the feature values.(3) Preprocess of the image, including image normalization size, format unification, denoising and image enhancement, which ignores the background of image. Then calculate the target area, elongation, Hu moments and some characteristic values. Finally find similar images by similarity matching and achieve the shape-based image retrieval.(4) Design retrieval system, which consists of image input module, feature extraction module and the image retrieval module. The introduction of relevance feedback improves retrieval efficiency.(5) Combine the two methods based on color features and shape features to retrieve image database. Match the similarity of images according to Hausdorff distance of image feature vector, and output the retrieval results to the user in order of the similarity through image display module.Experimental results show the method used in this paper can effectively achieve the goal of searching cars and other targets.