一种视频序列中的运动对象自动分割算法

A New Method for Moving Object Segmentation Automatically in Video Sequence

提出了一种自动、准确的运动对象分割算法。首先通过直方图拟合获得准确的背景噪声方差,克服了以往只能依据经验设定背景噪声方差的缺点,并使用显著性测试技术有效地对帧差图进行二值化,确定出运动区域。然后进行形态学和对称差分处理消除噪声及显露背景,获得初始运动对象。但由于分割结果不够精确,再使用梯度向量流场作为外力的改进活动轮廓算法得到运动对象精确轮廓。实验结果表明,该方法能够得到运动对象精确的轮廓,并且具有调整参数少,抗干扰能力强,可并行处理等优点。

An automatic and accurate moving object segmentation algorithm is proposed in this paper. Firstly, to overcome the shortcoming of setting the value of background noise variance by experience, we estimate the value by histogram fitting. At the same time, we use the significance test to threshold the difference image and extract the moving areas. Then, the residual noise and the uncovered background due to the motions of objects are fast and effectively eliminated by morphological operations and intersection of two symmetrical moving areas. Thus the initial moving object and its contour can be obtained. However, the segmentation result is not very accurate. To solve this problem, an improved active contour which uses the gradient vector as the external force guides the initial contour moving to the actual video object contour. Experiment results testify that the proposed algorithm not only can obtain the closed and accurate video object contour but also has the properties of few parameters, robust to noise, and parallel processing