Research On The Techniques Of Texture And Shape Based Image Retrieval

Multi-feature based image retrieval is an important research issue of content-based image retrieval. This paper sets out with the shape feature and texture feature, the current status and future development trend of content-based image retrieval are summarized,and the fundamental, key techniques and performance evaluation of CBIR are analyzed and summarized. One image retrieval approaches are proposed based on the human visual feature. The texture primitive of images is defined by the principles of BTC. According to the distribution of these primitives, a texture primitive co-occurrence matrix is introduced and a few of significant values are extracted to describe the texture feature. At the same time, primitive histogram is presented as the shape feature of images. Then, integrating the two features to image retrieval,we design and complete a prototype of a texture and shape-based image database retrieval system.Many kinds of high-dimensional index methods proposed recently have been investigated and compared. The VA-file approach can work well in the high-dimensionality cases, and the emulation experiments on the VA-file have been made by using MATLAB. The result of the experiments show that VA-file approach can improve the query capability by reducing I/O cost of system in high-dimensional index structure.