基于图像融合的运动前景检测方法

提出了一种有效的运动前景检测方法。该方法根据图像融合思想，将背景帧与监控视频的当前帧在R，G和B颜色通道分别进行融合，形成包含背景帧和当前帧视觉信息的单一融合图像。之后根据背景区域与前景运动目标在饱和度上存在较大差异的现象，使用大津算法分割融合图像的饱和度分量图，形成运动前景二值图。经形态学处理后，形成了目标区域较完整、背景干净的运动前景检测图。实验结果显示，该算法具有较好的前景检测性能，解决了背景减法过分依赖背景帧的缺陷。

Method for moving foreground detection based on image fusion

An efficient moving foreground detection method presents in this paper. The background frame and current frame in monitoring vedio are fuzed in the R, G, and B color channels according to the image fusion idea to form a single frame ineluding visual information of the background frame and current frame. The Otsu segmentation algorithm is used to get a moving foreground binary image by segmenting the saturation component of the fused image to eliminate the phenomenon that the back-ground region and foreground moving target have a big difference on the saturation. The moving foreground detection diagram with a more complete target area and more clean background can be formed by morphological processing. The experimental resuits show that the algorithm has better foreground detection performance and can eliminate the defect caused by background subtraction algorithm which relies on background frame too much.