Content-based Video Motion Object Extraction

To meet the multimedia communications particularly video conference and video telephone application demands, new video coding standards such as MPEG-4, MPEG-7, etc. came into being. Meanwhile, the setting of new video coding standards also greatly promotes the storage and communication technology of multimedia information.Video object segmentation is the key technology of the second generation video coding standards with representative MPEG-4 and has become a research focus in the field of digital communications and even the computer vision in recent years. Although a lot of work has been done in this field, video object segmentation still stays in the stage of moving object segmentation and still needs much further research.This paper elaborates the related technology of the moving object segmentation. Firstly, some relevant information of the video object segmentation has been introduced such as the generation, development and classification and so on. The second, some basic knowledge of video object segmentation have been introduced and some segmentation techniques of spatial domain and temporal domain have been analyzed. Then the global motion estimation and compensation techniques of video sequence with moving background have been discussed in detail. Finally, the change detection technique of automatic segmentation has been studied.After reading a lot of domestic and international relevant literatures, the following jobs have been done:(1) Studied the cause of moving background of video sequence, common moving model of camera, the theory and mathematical model of global motion estimation. A new method of parameter estimation is proposed in the article based on the analysis of the characteristic of global motion and local motion. The method classifies the motion vectors from block matching directly and then finds the reliable background vectors which will be used to estimate the global motion parameters. A lot of experiments prove that the method is efficient to get satisfactory compensation results.(2)Deeply studied the algorithm-difference between neighboring frames, its'improved algorithm-intersection between forward difference and backward difference and researched on the other automatic segmentation algorithms. During the study, the author finds that one of the causes that affect the segmentation speed is that the computation amount of the binarization of difference image is too large. In order to improve the speed, the article chooses Ostu algorithm for binarization, and replaces the difference image to the summation of several difference images. The method strengthens the difference between the background class and foreground class and can get an ideal global threshold using Ostu. Lots of experiments prove that exact moving objects can be extracted by the method assisting with other digital image processing methods.