基于混合概率背景模型的视频分割方法

提出一种新的基于混合概率模型的背景建模方法,用于视频中前景物体的检测与分割。主要利用两个概率模型：隐马尔可夫模型和概率图模型建立一个混合的贝叶斯网概率模型,对视频输入中背景变化的时间和空间局部相关性(同现性)进行学习。在建立正确模型参数的基础上,贝叶斯信念传播算法根据图像输入预测当前背景状态的后验分布,并根据预测得到的背景状态对输入图像进行分割。实验结果验证了该方法的有效性和在复杂背景变化下的鲁棒性。

Video segmentation method based on mixture probabilistic background model

Aunified framework was proposed for detecting and segmenting foreground objects in complex scenes involving swaying trees, moving shadows and ocean waves. A mixture probabilistic model of hidden Markov modelsand probabilistic graphic model was presented to model the variation of background. In this model, the color of each pixel from backgrounds was regarded as a random variable. The property that background variations at neighboring pixels had strong correlation, also known as "co-occurrence", was used to initialize and update the mixture model online. Bayesian belief propagation algorithm was used to efficiently calculate the maximum of the posterior probability of background with input data. The experimental results for real video demonstrate the effectiveness and robust of our method.