A Method Of Image Retrieval Based On Texture Feature Extraction

With the Multimedia information widely used, and the promotion of the research of database system and computer vision, content-based image retrieval(CBIR) is becoming a hot research area. How to analyze, store and retrieve these images is an imperative problem. CBIR technology can effectively resolve them and it also becomes a hotspot in study. CBIR is different from text-based image retrieval. In fact, CBIR is a system which can realize organizing and retrieving images automatically and intelligently through extracting some features from images and finding the most similar images which have the most similar feature with the indexed image. Feature extracting and matching is the key point of deciding the result of the CBIR system.We compared several feature extracting methods which are color features, texture features and shape features. Since texture feature can describe the properties of smooth, sparse, and order of images, this paper decides to do image retrieval based on texture features.At first, the paper introduces the current research situation of CBIR both at home and abroad.According to the rationale of CBIR,it discusses the general frame,key technique,inquiry ways and application fields.Then describing the definition of texture and texture feature in detail;introduce some usual methods to analyze the texture, including gray histogram, edge histogram, gray concurrence matrix, Tamura texture; particularly describe Gabor filter, Gabor Wavelet and the method of Gaussian normalization, then try the method of extracting texture with gray concurrence matrix and Gabor filters.In chapter 4,the paper introduces the geometric matrix model and set theory model, propose two methods of Comparability measurement structure with multi-feature and the relevance feedback and the similarity measurement.At last, designing an image retrieval system and describing each module's function in detail. To appraise the algorithm with a sorting assessment method and the average retrieval rate, carrying out the emulational experiments in the paper.