The Research Of Object Extraction Algorithm Based On Markov Energy Clustering

As the important information in the telemedicine system, video image provides the important gist for real time long-range medical diagnosis and wardship. Video information and its processing technology have made much progress. Because of the enormous data, it is quite difficult to be stored and real time transported. So the problem is hindering the application of telemedicine technology development. Hence processing and analysis for video image gets feature information, which is vital to storage,transmit and search the telemedicine video image.Robust segmentation and tracking is an active field of computer vision. Segmentation of moving objects is to find differential objects in video image sequences and track them. This study is the foundation of video retrieving based on content and video coding based on objects etc. However, video image includes video object multiformity and environmental complexity, it brings huge difficulty to video object segmentation and tracking. Hence video object segmentation and tracking is regard as a classical problem in computer vision and image processing field. The study of video object segmentation and tracking gains some evolvements, but now it still has no a automated segmentation and tracking algorithm for all video images. Therefore, this paper pays attention to the special algorithm practicability and develops some key questions study. The study includes two points: one is to exactly segment video objects in video sequences, the other is to make video objects tracking.Firstly, in the research on arithmetic of standard FCM clustering, we extend FCM by increasing different spatial information. As a result of video image spatial correlation, we can use the correlation for extending FCM clustering algorithm. It will acquire more accord with man perceptually segmentation results. Experiment proves that joint spatial segmentation method improves video object segmentation performance.Secondly, by deeply studying all kinds of statistic segmentation algorithms, the paper ameliorates a video object segmentation algorithm based on MFT using MRF-MAP. This algorithm begins with a simply foreground image derived from the difference between video sequence two frames. Computing the energy function of MRF is constructed by applying the MFT. The MRF-MAP which is minimum energy function, is adopted to obtain the label field and extract the video object. The Experiment results show that this segmentation method is sufficient thinking of image local spatial correlation. The model can not only extract single moving object from the simple background but also preferable extract the moving objects from the complicated background by using Matlab.Thirdly, combine MRF model with FCM clustering, the paper presents a markov energy clustering algorithm. Firstly, look for initialization label field by fuzzy clustering, then by the means of MRF to compute every site characteristic energy, through energy global clustering and gain a new segmentation label field, recalculate every site characteristic energy based on the new segmentation label field, then carry through clustering segmentation, continue circulating segmentation, finally gain optimize segmentation result. Through fuzzy clustering algorithm bring to bear MRF model spatial restriction field, consequently enhancing algorithm noise immunity. The Experiment results show that the paper presents method can acquire better video object form information, and can precise extract different video sequences'objects.Lastly, the paper presents a simply video object tracking algorithm based on minimum rectangle frame. By drawing a frame round of the former frame moving object, computing the least restriction frame,moving speed and direction, we can quickly track next frame the site of moving object, it can effective reduce calculation and complexity of video object segmentation. The Experiment results show that this method can superduper track video object and gain the edge of approaching video object.In a word, the paper presents a markov energy clustering algorithm. Simultaneity, it takes standard MPEG-4 sequence image for research object, many experiment results validate that this method could effectively reduce the object segmentation effect of noise and object non-rigid moving, it can acquire perfectly video object segmentation result. As real time tracking video object, it can effective reduce calculation and complexity of video object segmentation, and ensure image sequence segmentation exactly and real time. Although this method makes some efforts in aspect of universal, we still need to do more work in deeper research.