Complex Scene Constraints Area Filling Technology Research And Realization

Video features extraction and filtering is an import part in image processing, video processing and pattern recognition field. It has been widely used both in military and civil fields. With the development of Internet, people can retrieve all kinds of material on the world-wide-web. Pornography content is harmful for under-age net surfer, this is a serious social problem. Besides legislation way, it calls for effective techniques to keep children from accessing objectionable content. There are three main content-based methods: URL-based filter, text-based filter and image content-based filter. The third method is a promising, thorough and effective way for it has universal adaptation and the capability of coping with the changing circumstances.Content-based video sensitive information filtering and video-recreating is the work platform of my work. We try to gain visual acceptable and less artificial video through removing the whole sensitive object and reserving the integrity of other information especially the video structure. The main work of this paper is motion detection in video frames stream and the region filling after sensitive object removal. The main works of this paper are as follows:1. This paper introduces the Content-based video sensitive information filtering and video-recreating platform through analysis of the three basic methods used in web content filtering field.2. This paper gives a brief description of the structure of video and the classification of video shots, and then generalizes the common methods to extract features and build metrics for computing the discontinuity values. Also this paper reviews the algorithms for detecting the cut and gradual shot changes and their advantages and drawbacks.3. This paper summarizes the current technologies used in motion detection task and proposes a pixel-classification-based background recreation approach for motion detection task.4. This paper introduces some review on the basic principles related to image completion including the"best guess"principle and Bayesian framework and basic theory of variation. This paper emphasizes on variation-based model for little scale region filling task and texture synthesis approach for large scale region filling task specially.5. This paper introduces variable parameter based variation model for little scale region filling task by analyzing the deficiencies of total variation model, and also presents the numerical algorithms.6. This chapter introduces an improved image completion model through using local gradient to determine template window's size adaptively by analyzing the deficiencies of current image completion models.