基于差分交集的视频对象分割与跟踪算法

Video object segmentation and tracking algorithm based on difference and intersection

视频对象分割算法的性能好坏将直接影响MPEG 4编码产品的质量。连续两次差分后自适应处理,对差分图像取交集获得运动对象的边界,形态学处理后最终获取运动目标。基于改进的Hausdorff距离度量法对后续帧中视频对象进行跟踪。实验结果证明,该方法能够从背景不变的图像序列中较好的提取出运动对象,具有较强的鲁棒性。

The performance of video object segmentation algorithm has the direct effect on quality of MPEG-4 products. Make twice FD for three continuous sequence frames and manage adaptively, then take an intersection for two difference images and obtain the motion object's edge. Last pick-up the moving object is obtained by mathematic morphologic method. Then the following frames are tracked based on the improved Hausdorff distance. Experimental results demonstrate that this method can effectively separate the video object from the sequence with static background and has the strong robustness.