The Research On Compression And Coding Based On Video Object

MPEG-4 is an object-based and very open video compression and coding standard. It allows competition and improvement for the video segmentation, rate control and coding technology and so on. The key technologies of MPEG-4 are becoming the hotspot of study and development.Based on the research of the basic idea of MPEG-4 technology and the key algorithm, this paper emphasizes on the study of video objects segmentation and motion estimation and compensation technologies of MPEG-4.In the video segmentation part, a new segmentation approach based on spatio-temporal information is proposed. The new algorithm firstly selects four rank moment method to segment temporally to obtain the coarse mask of video object. Then it uses the spatial information to fine the result to improve the coarse boundary after temporal segmentation. The fining process proposes a cost function about the pictures transparency, according to the coarse mask and using the local smoothness and correlation property of the video sequence gray. By minimizing this cost function, the transparency of each pixel can be solved and the gray mask of video object can be obtained. Because of using both spatial and temporal information, the new algorithm improves the fault of the old method that the extracted outline is blur. The experiment shows that the result of the new algorithm is very accuracy.In the motion estimation and compensation part, Through research the classic motion estimation algorithm and the cross-diamond search algorithm, proposes an adaptive rood pattern search using the characteristic of different kinds of block with thinking about the features of the variable block-size motion estimation and the spatial-temporal correlation of motion vector. The new algorithm predicts the motion of the current block firstly and then chooses the pattern adaptively according the prediction result. It solves the disadvantage of single pattern in searching vertical moving objects and enables us to add the number of search points more efficiently and thus improves the search speed. Meanwhile, the new algorithm also adopts the advanced-stop criteria and linear search technology. Simulation Results show that the new algorithm can efficiently reduce the search time.