Research Of Video Object Segmentation Algorithm Based On Change Detection Mask

Nowadays, the video encoding method is developing from the first generation (Marco Block-based) to second generation (Content-based and Object-based). Object -Based and Content-Based, as a new method, deeply affect the digital video encoding and retrieval. But first of all, we must realize the function of separating the video object from background.This thesis studied the existing video object segmentation algorithm and analysis their excellences and weaknesses. Some improvement and new algorithm has been presented. The important ones are as follows:①On the condition of static background. The paper use Symmetrical DFD algorithm to get the counter of moving object, and a simple adaptive binary threshold approach and counter fold method has been brought forward. Finely solve the unveiled and sheltered district of background.②The paper presents a new algorithm Combined with Canny edge detection and Symmetrical DFD. The experiment indicates this algorithm has a good performance on segment accuracy.③Study the global motion estimation and compensation technology , presents a new video object segement algorithm under golbal motion, And actually segment the object from the moving background.Experimental results show that the algorithm can accurately and robustly segment the video object. The algorithm has a low complexity, and don't need manual intervention, suitable for use in real-time.