Content-based Image Retrieval Technology

With the quick development of internet, as well as the wide application of multi-media, more and more images are produced. It becomes an urgent issue that how to find the content people need from the huge image database and visual information, it is the key that whether can effectively organize, manage and retrieve these huge image database. Therefore, Content-based image retrieval(CBIR) technology emerges in this background, and quickly becomes a hot research topic in image processing domain. This paper systematically analysis and researches CBIR technology.This paper roundly grasps the background knowledge and application of CBIR, understands the design idea and classical algorithm of famous image retrieval systems at home and abroad, in order to lay the foundation for the development and deep research of this topic.In the field of image retrieval based on color feature, according to different applicable scope of different color space models, chooses suitable color space model for this research, then extracts and quantizes color features, in order to reduce the computational complexity. On the basis of traditional global color histogram and image uniform block, this paper proposes a method which is based on improving blocked cumulative histogram and the retrieval experiment has a better result.This paper roundly analysis the application of texture features in image retrieval, introduces three different texture analysis methods, includes statistical analysis method, structure analysis method and spectrum analysis method. It analysis the traditional gray symbiotic matrix method in detail, focuses on local binary mode (LBP) theory, and proposes LBPV algorithm based on LBP theory. Experiment shows that the proposed algorithm is superior.This paper analysis and chooses different similarity measure tools of image features depend on different image retrieval methods, so each method has the best retrieval result and the highest average hit accuracy, and all the experiment results are compared and analyzed.