一种基于混合高斯模型的运动阴影去除算法

Moving cast shadow elimination algorithm in Gaussian mixture model

为更好消除光照条件下运动物体阴影,在混合高斯模型(GMM)识别前景目标基础上,提出一种基于RGB颜色模型的阴影消除方法。在RGB颜色空间中,首先用混合高斯模型获取带有阴影的运动前景像素,利用该模型参数,估计该像素点的阴影模型误差,然后建立RGB阴影模型;由于阴影像素点像素值符合阴影模型,将该阴影模型作为混合高斯阴影建模(GMSM)的匹配条件,最终确定当前像素点是否为阴影像素。实验表明,算法在不同场景中都能使阴影去除达到满意效果,同时满足常规视频序列的实时性要求。

In order to provide great elimination of moving objects shadow in light conditions,RGB shadow model is proposed using Gaussian mixture shadow elimination algorithm,based on Gaussian mixture model( GMM) identifying prospects. In RGB color space,the shadow model of pixel error is estimated by GMM parameters previously and a RGB shadow model is established. This shadow model is a matching condition for modeling Gaussian mixture shadow model. Therefore,current pixel is determined whether it is shaded or not ultimately. Experimental results demonstrate that the proposed algorithm achieves satisfactory results for shadow removal in different scenarios,with real-time requirements of conventional video sequence.