Content-Based Thangka Image Retrieval Technology Research

With the rapid development of computer and Internet technology, the data source of image becomes larger and larger, which leads the research focus all over the world to be content-based image retrieval technology, with which people can retrieve targeted image from plenty of images quickly and correctly. Through content-based image retrieval technology, retrieval system usually extracts different characteristics according to different images. Therefore, the key technology is to extract image characteristics. The object of this study is Thangka, the digital study of which plays an important role in protecting Thangka culture heritage. In order to obtain satisfying result of Thangka image retrieval, this thesis makes a detailed study.Firstly, this thesis systematically studies extraction technology of visual characters about image's low-rise. On the basis of that, the typical extraction methods of characteristics of color and shape are respectively studied. Retrieval system for color-based and shape-based Thangka image are designed and realized. In addition, during the process of extracting color characters, an improved method called local accumulative histogram based on HSV is come up with combined with traditional local accumulative histogram and Thangka image's characteristics. Through many experiments, it shows that the color characteristics seem to be more important than the shape characteristics and the improved method called HSV-based local accumulative histogram is superior to traditional local accumulative histogram. Secondly, according to separate characteristics of color and shape, those two experimental retrieval systems are used to analyze the result Thangka image retrieval system. Then a comprehensive color-based and shape-based experimental retrieval system is designed and realized. After the experiment result analyzed, it is found that with comprehensive kinds of characteristics the efficiency of retrieving will be enhanced.