论文部分内容阅读
提出一种修正分水岭算法，用于视频自动分割，包括初始分割和时域跟踪两方面。该算法可以有效跟踪快速运动的物体，并检测出新物体的出现和原物体的消失。此外分水岭变换和快速运动估计的应用大大减少了计算工作量，提高了效率。模拟结果表明，这一技术对于图像中包含有快速运动物体，或有物体的出现或消失的视频序列可以有效地进行分割。
A modified watershed algorithm is proposed for automatic segmentation of video, including initial segmentation and time-domain tracking. The algorithm can effectively track fast-moving objects and detect the appearance of new objects and disappearance of original objects. In addition, the application of watershed transformation and fast motion estimation greatly reduces the computational workload and improves the efficiency. The simulation results show that this technique can effectively segment the video sequence that contains fast moving objects in the image, or the appearance or disappearance of objects.