一种改进的基于码本和高斯混合模型的视频背景分离

An improved video background separation based on codebook and Gaussian mixture model

文章提出一种基于改进的码本(CB)和高斯混合模型(GMM)的视频背景分离方法。该方法是以自适应的高斯混合模型背景为基础,为每个颜色像素构建混合高斯背景模型,可以对视频帧中每个像素的高斯分布数进行动态控制,并且通过CB(Codebook)算法得到每个像素的时间序列模型,从而对高斯分布的各参数进行学习。实验结果表明,该方法在背景分离的精确度和处理时间上都表现出优异的性能,此外还具有良好的适用性,对复杂场景的变化,可以有效快速地分离视频的前景和背景。

This paper proposes a video background separation method based on improved Codebook （ CB ） and Gauss mixture model （ GMM ）.This method is based on the adaptive Gauss mixture model background,and constructs a hybrid Gauss background model for each pixel color,which can dynamically control the Gauss distribution number of each pixel in the video frame,and through CB （ Codebook） algorithm to obtain each pixel of the time series model,and study on the parameters of Gauss distribution. The experimental results showthat the proposed method in accuracy and time background separation have shown excellent performance,also has good applicability,ccan effectively separate the foreground and background from a video.