为了使GVF-Snake具备快速跟踪运动目标的能力,提出采用帧差运动补偿和GVF—Snake结合的跟踪算法.采用二次帧差法对视频图像进行处理,自动得到初始轮廓解决GVF—Snake模型初始轮廓确定的问题.对于输入视频图像在每一次迭代中重新计算整幅图像的GVF场时间消耗大的现象,采用分块计算,减少了计算量,使GVF—Snake跟踪达到接近实时的要求.在遮挡方面,采用前一帧差运动补偿的方法.实验表明,此跟踪算法具有较好的性能.
In order to enable GVF-Snake track the moving objects quickly, a compensation method with frame-distance is proposed to join in the GVF-Snake to track the moving objects. The frame-difference method of the second order is presented to obtain the initial contour of the input video image which deals with the initialization problem of snake model. The computational load of computing the whole image GVF field would be huge in every step when getting the video image sequence. However, it would be reduced by just computing the GVF field in the sub-area of the image interested. A compensation method based on the previous frame-distance is presented to deal with the problem of occlusion. The results of experiments show that the proposed method has good performance.