Research And Implementation Of Image Retrieval Based On Color And Texture

In 1990s,the content-based image retrieval(CBIR) technology is proposed and becoming a new hotspot in less than twenty years.The technology of CBIR analyzes image content and extracts image features directly,such as color,texture and shape of the image,which are used to built index and then retrieval.It overcomes the limitations of traditional image database retrieval based on keywords or descriptive text for its low effectiveness and ambiguity.The research of CBIR will provide powerful support for the development of multimedia database.Meanwhile,the technology of CBIR is of significant importance to the development of next generation Internet,information query and content-based video retrieval.Considering the current status and problems of the CBIR,the paper has researched on the following aspects:1) This paper expounded and summed up systematically the key techniques of CBIR,including content-based feature description and feature extraction,calculation of similarity distance and evaluation of CBIR performance etc.2) Through the research and analysis of different color space and color features, the paper adopted HSV color space and proper quantization to extract color histogram and color moments.As the distribution of colors can't be embodied in the above two algorithms,blocked histogram is used to enhance the color information of image's special location and improved retrieval efficiency greatly.3) The paper compared two retrieval algorithms based on texture features:gray level co-occurrence matrix and gabor wavelet.Based on single feature,multi-feature with both color feature and texture feature for image retrieval is proposed.They are weighted respectively.In this paper,multi-feature is composed of color histogram in HSV space and gray level co-occurrence matrix.The experiment results show that compared with single feature,retrieval with multi-feature accords with people's visual feeling better and also has better retrieval results.4) Corel images and 500 Brodatz images are from standard image database and used in the experiments.The paper researched advantages,disadvantages and applicability of various algorithms through their recall and precision.Then,a image retrieval system based on content is designed and realized.In the system,users can make retrieval by various single features and also multi-feature.The feature database is built.For the retrieval with multi-feature,relative weight can be decided by specific image.