针对目前运动对象分割不完整,以及存在阴影和鬼影对运动目标分割的影响,提出了一种基于复杂背景下的运动目标分割与阴影消除方法.首先利用高斯混合模型进行初始背景建模并提取初始前景对象,将当前视频帧和背景模型进行差分运算,且通过多尺度小波变换时空域特征,将多尺度分析和图像分割相结合,压制阴影并消除鬼影对运动目标分割的影响而得到前景对象.通过实验对比,所提方法能有效地从复杂背景视频图像中提取运动目标且具有强的鲁棒性.%Aiming some existing limits in foreground objects segmentation such as incomplete segmentation of moving targets, moving shadow and ghosts, a novel segmentation method and shadows elimination from a complex background is proposed. Firstly, a Gaussian mixture model ( GMM) is adopted to construct background model and extract some foregrounds. Background subtraction is performed between the current frame and the previous constructed background model. According to some characteristics of multi-scale wavelet transform in spatial and temporal fields,multi-scale analysis is combined with image segmentation to suppress shadows and eliminate ghost from the foreground in the video. Experimental results show that the proposal can effectively segment the moving targets from video with a complex background and eliminate shadows and ghosts by comparisons.
展开▼
