The Research Of Integrated Multi-features And Feedback For Image Retrieval

With the popularization and development of the Internet ,large numbers of image data are produced fastly. The traditional text-based retrieval method can no longer meet the needs of image retrieval.Image intrieval has become a hotspot for the present application field of image.It combines many fields of knowledge,such as image recognition, database technology,computer vision artifical intelligence.The key issue of CBIR is extracting features from low-layer image features.After the paper researched on the background and meaning of the CBIR .We introduce every classification algorithms and key technology.The paper researched on the technology of the CBIR ,the main content is following:On the aspect of feature extraction,three features extraction methods are analyzed separately, like color, texture and shape features extraction;On the extraction of image feature,the paper researches low-layer feature extraction algorithm based on color,texture,shape.We use HSV color space during color feature extraction.In order to overcome the lack of spatial information after dividing the image and then quantified on the HSV color space.Finally we combine with cumulative histogram to extract features.In the extraction of texture features,because when extracting texture feature on the co-occurrence matrix,it has little amount of calculation,faster speed of feature extraction.and it has little feature vector dimention.Therefore we use the method in the paper. On the aspect of feature extraction,we researched the various algorithms of shape descriptor ,we provided a method based on shape feature with integrating on reigon and conlour's features. A shape feature descriptor based on zernike moment descriptor on the circle partition with different distance is providedAnd we use the canny edge detector to extracts the shape features.After processed the edge's points,we compute the distance between edge's points and center of gravity and get mean and variance.We can use the edge and contour's feature of image when we use the two methods at the same time.The method can accept more shape information than traditional method and achieve better search results.Finally,on the rasearching of the above algorithm,an experiment system of the CBIR combined color,texture,shape is designed.In order to reduce the gap of underlying features and high-level concepts,we adjust the weight of multi-features according to the feedback information form the user's subjective evaluation.