Research On Some Problems About Still Image Retrieval And Digital Video Parsing

In this thesis several questions about still image retrieval and digital video parsing are investigated. With the quick development of information processing technology and the increment of applications based in World Wide Web and database technology, it has been a urgent problem to provide high level, abstract and structured store and retrieval methods to large amount of image and digital video data, so that one can simply, quickly and effectively use these data. The research work of this thesis focuses on image feature abstraction and digital video parsing. On the basis of analyzing related literatures, some impressive contributions are list as follows.1. To the problems in content based image retrieval, after generalizing the operation object and operator, a new extraordinary Radon transform method is proposed and three kinds of transform are defined. The first one of three kinds of transform are the pixel Lebesque integral based projection Radon transform, region Lebesque integral based projection Radon transform and region set operation based projection Radon transform. In this thesis three kinds of transform are unified in one theory system. On the basis of the definition of the extraordinary Radon transform, the theory of rebuilding image from extraordinary Radon transform result is proposed, and some category such as image structure notation, topology of image structure notation and topological space, homeomorphism in topological space, equivalence of homeomorphism relation, topological constant of image structure notation. The formal analyze about the relation between the similarity of image structure and the equivalence of homeomorphism of image structure notation is...