Research And Implement Of Texture-based Image Retrieval

According to the development of computer network and multimedia technology, more and more image information appear in people's lives. How to find out the picture we need in huge amounts of image data became a research focus. Therefore content-based image retrieval technology appeared. It is different from traditional text-based search. It is actually a fuzzy inquiry. By extracting certain features of the image it finds out similar images in feature space.The texture-based image retrieval technology is deeply studied in this paper. Firstly, the basic theory and frame of content-based image retrieval technology are introduced. Then the extraction of texture feature is studied deeply, focusing on the research of Gray Level Co-occurrence Matrix and the 3D Co-occurrence Matrix which including color information. Afterwards, the research of key technologies on content-based image retrieval system is done, including similarity measure and relevance feedback. Finally, designs and implements a texture-based image retrieval system and evaluates the system's performance and proposes future research directions.In the part of system implementation, the paper used a method to sort result images based on similarity as calculating. And a relevance feedback mechanism is added to the system, by adjusting the weight to search again then sort and show images using the new rules. Through the experiments of Corel image library, the results showed that by combine 3D Co-occurrence Matrix method and the relevance feedback mechanism designed in this paper can make the search results more meet users' needs.