Abstract: This paper proposes a shadow removal algorithm for moving pedestrian based on weak thresholds segmentation. The thresholds for segmentation are classified into weak and strong. The foreground is extracted by Gaussian Mixture Model(GMM). The weak thresholds are applied in color space and gradient space respectively to get partial objects, which are merged next. After the neighbour fixing, the complete object with no shadow pixels is obtained. Experimental results prove the validity of the algorithm.

Key words: weak thresholds segmentation, Gaussian Mixture Model(GMM), shadow removal, color space, gradient space

摘要： 提出一种基于弱阈值分割的运动人体阴影去除算法。从阈值选取的角度出发，定义强阈值与弱阈值，使用混合高斯模型获取运动前景，在颜色空间与梯度空间中提取弱阈值目标，融合提取出的多个目标区域，经过邻域处理后得到不包含阴影点的运动人体目标。实验结果证明了该算法的有效性。

关键词: 弱阈值分割, 混合高斯模型, 阴影去除, 颜色空间, 梯度空间

CLC Number: 

• TP391