基于水平集的多运动目标检测和分割

针对视频图像的运动分割问题，提出了一种基于水平集方法的多运动目标检测和分割新方法.通过一种基于帧间差分的算法,自动提取初始背景图像，并使用相减法，检测出当前图像中的运动像素.定义了一种新的基于差分图像的局部梯度、目标的方差和背景的方差的速度函数，得到了改进的分割曲线的演化方程，分割出不同的运动目标.在水平集的求解过程，设定了控制演化曲线最终停止在目标边界上的条件，得到了运动目标的边界.实验结果表明，与其他传统方法相比，该运动目标检测和分割方法更有效和具有更好的鲁棒性，能够正确地提取运动目标边界.

Detecting and segmenting multiple moving objects using level-set method

Aimed at the motion segmentation problem in video sequence,a novel approach for detecting and segmenting multiple moving objects was presented based on level-set method.The frame difference algorithm was used for dealing with the problem of initial background image extracting automatically,and different moving pixels in current image were detected through a subtracting technique.After a novel speed function based on local gradients of differencing image,object variances and background variances was defined,an improved evolution equation for segmentation curve was obtained,and then different moving objects were segmented.During the solution of the level-set equation,a condition controlling the curve evolution to stop on the object boundaries was set,and the moving object boundaries were achieved.Experimental results show that the proposed approach for detecting and segmenting multiple moving objects is much more robust and powerful than other traditional methods,and can exactly extract the moving object boundaries.