基于区域相关的核函数背景建模算法

传统的背景建模方法都是单独为每个像素建立背景模型,这样就没有考虑像素之间实际存在的相关性,使得在进行前景和背景分割时出现不该出现的空洞和噪声点。为了解决这一问题,提出了一种基于像素邻域信息的核密度估计背景建模算法;给出了相邻像素亮度值分布模型间的相关性描述,并将该相关性引入到背景建模中,得到包含时一空信息的背景模型,并提出背景模型的更新策略;实验结果表明,算法能有效去除前景区域内的空洞和背景区域内的噪声点,能快速准确地更新背景模型,能进行复杂场景的背景建模。

Region-Based Correlation Method of Kernel Background Modeling

Because of traditional background modeling method do not take into account the correlation between pixels,there will be holes and much noise when the image segmentation is done.In order to overcome this problem,a method of kernel based background modeling with neighbor- hood information is proposed.The correlation of intensity distribution models between proximal spatial pixels is described,and this correlation is used in background modeling.Then,the background model included temporal-spatial features is produced,also,the strategy of background model update is presented.Massive experiments indicate that the method can eliminate holes in foreground regions and reduce noise in background regions,can up- date background model rapidly and accurately,and can estimate the background model of complex scene.