中文摘要：

针对现有目标轮廓提取方法存在收敛速度慢、效率较低以及对初始位置和噪声都很敏感等问题,提出了一种差分相乘与参数多水平集主成分Chan-Vese模型相结合的新模型.该方法首先利用4帧相邻图像进行差分相乘,抑制绝大部分的背景边缘,再进行滤波预处理,最后通过多水平集与Chan-Vese模型结合的改进模型提取运动目标轮廓.对大量视频图像进行实验分析后的结果表明,在具有多个目标的视频图像下,该方法能更加快速准确地提取出每个完整的运动目标轮廓,较好地解决了现有方法在多个运动目标轮廓提取以及轮廓凹陷上的缺陷.

英文摘要：

Aimed at the problem of slow convergence speed and low efficiency for moving target contour extraction,a kind of new algorithm based on difference multiplication and multi-level set with Chan-Vese model are put forward,dealing with the problem of sensitive to the initial position and noise of C-V model.The method uses four adjacent images frames for differential multiply to inhibit and filter most of the background edge,and then extracts moving object contour through multi-level set with C-V model.Experiment results show that this method not only overcomes the difficulty of multi-objects recognition,but also accurately extracts moving target contour more quickly.