Technology Research, Content-based Image Retrieval

With the rapid development of the computer technique, communication technique, multimedia technique and network technique, there are more and more resources of figure image. How to organize and utilize the image source that become more and more massive and how to search the image fast and effectively, has become an earnestly problem to resolve in the figure image domain. All the problems have promoted the research and development of image retrieval technology.Traditional image retrieval technology based on text hasn't been satisfied with the retrieval demand of immensity image source, because the technology need a great deal of manual work and have low nicety in search-matching. CBIR, which combined with image comprehension, pattern recognition, and other techniques, can provide users with more efficient measures in image retrieval. So CBIR has become hot spot in image retrieval domain. Based on the current research results, how to extract image content features, and how to improve the accuracy of image retrieval, both become the problem that deserves more study.This thesis mostly discusses some issues in CBIR. Firstly, the thesis analyzes the approach of image retrieval engine in network and the present situation of development and the existed difficulty of CBIR.Then, the thesis analyzes the expression and extraction techniques of image color feature from color histogram, color moments, color coherence vector and color correlograms. And analyzes contrast the SAR texture feature, Tamura texture feature and GABOR texture feature. The paper analyzes the shape feature form Fourier-based shape descriptors and moment invariants in CBIR. And analyzes contrast the arithmetic of image division, such as the arithmetic based on verge and the arithmetic based on area.Then discusses some related techniques in image retrieval, such as multi-dimensional indexing technique, knowledge and semantic technique and relevance feedback etc, the thesis also analyzes the similarity measures and appraises technique.At last, the thesis designs the experiment image retrieval system based on color feature. The system realizes the image retrieval based on key word and content. Finally, The thesis compared the effect of two methods.The fulfillment of this thesis realizes the effective expression, extraction and retrieval of image features. It provides users with some theory methods and practice techniques to find the interesting images quickly from the vast image resources, and at the same time the content-based image retrieval experiment system also provide reference for the commerce system designing in future.