Research And Implement Of Digital Content Retrieval Based On MPEG-7

In recent years, with rapid development of multimedia technology and computer networks, capacity of digital images, videos and audios all over the world is increasing at surprising speed. Both military and civil equipment are producing digital media materials which contain large quantities of useful information. However, disordered distribution of these materials worldwide makes it impossible for the useful information to be effectively visited and utilized. This requires a technology for fast and accurate search and visit of the materials, namely content-based retrieval technology.Content-based media retrieval technology is a hotspot of research at present. Many research organizations at home and abroad are carrying out investigations in this regard and have made certain achievements.MPEG-7 standard, which was brought forward by MPEG, emphasizes the description and definition of digital AV materials. MPEG-7 defines the AV materials with clear structure and expression, describe the AV content and contact the description with the content itself. This technique is very useful for content-based retrieval system.Currently, features used for retrieval mainly include color, texture and shape. Owing to different applications, different content-based retrieval systems emphasize different aspects in retrieval realization, but are generally originated from traditional computer visual and image processing technology, which means Content-based image retrieval is the combination of image processing and various computer visual technologies.In this thesis, thorough analysis and research has been done on image feature extraction, storage, and retrieval technology. The author proposed a framework of content-based image retrieval system which could use features such as color, texture, and contour to complete retrieval tasks. In addition, the author has implemented several key algorithms in the content-based image retrieval system.1. Survey the mainstream technology on content-based image retrieval systems to find the proper algorithms for image storage and retrieval.2. After analyzing each component of MPEG₇ standard carefully, especially the feature extraction part and XML, the author discussed the proposal of introducing algorithms in MPEG-7 into content-based image retrieval.3. As one of the key component in image retrieval, feature extraction has been introduced carefully in this thesis. The author described the symbols as well as the extracting algorithms of features like color, texture, and contour.4. To realize image retrieval, invariance of the image must be extracted for matching unchanged moment is a physical feature of the target, which will not change with the object in translation, rotation and change in proportion. The thesis extracts the unchanged moment feature of the image, uses Euclidean space to calculate distance for image matching to improve the effect, and obtains fairly good experiment result.5. Careful discussion has been done on the framework, functions and components of the retrieval system. Besides, the author provided description on some key issues concerning the implementation of the system, especially how to process the video segments.6. Upon the theoretic analysis introduced above, the author proposed a dual-mode storage- XML files and database--method for the feature storage, and analyzed the synchronous problems between these two modes.