背景重构是背景差分法运动检测的关键问题，为了快速地重构背景,提出了一种新的基于“背景点像素值出现频次最高”假设的背景重构算法，即先在观察时间内抽样若干帧作为训练图像序列，并用帧差法对图像进行处理得到大致运动区域，再利用区域生长法得到较完整的运动区域；然后将各抽样帧内运动区域内的点都标记为前景点，而将其余点标记为背景点；最后，统计每个背景点的不同像素值出现的频次，并选出频次最大的灰度值作为该像素点的背景灰度值。实验表明，该算法能快速有效地重构背景。

Background reconstruction is the key to moving detection. So a background reconstruction algorithm based on the hypothesis of background pixel with maximum probability is presented in this paper. First, we calculated the difference between sequential frames which are sampled from a section of video sequences, and get the full moving area by area growing, then label the pixels in moving area as moving points and others as background point. We calculate the mean of the intensity value of the background point in these frames [JP]and consider the mean of the intensity value as the background pixel intensity. The results of experiments show that this new algorithm can reconstruct the background quickly and successfully.