基于CNN和MRF的运动目标分割

针对动态图像序列中背景成像过程因各种因素而变化存在复杂性,提出了一种基于细胞神经网络(CNN)和马尔可夫随机场(MRF)的目标分割方法.首先根据细胞神经网络与马尔可夫随机场能量函数的相似性,将马尔可夫随机场的最大后验概率模型映射到细胞神经网络近邻系统模型中.然后建立图像每一像素点的邻域系统模型,并且构造相应的能量函数.为使能量函数达到快速收敛,再利用模拟退火算法实现能量函数的最小值,以达到对运动目标的提取.由于CNN是由局部互连的细胞组成,因此易于用VLSI实现.实验的结果表明,该方法能够有效地抑制图像的噪声,对于运动目标的提取有较好的分割效果.

Motion Object Segmentation Based on CNN and MRF

This paper proposes a novel approach using Markov Random Field and cellular neural networks for segmenting moving objects from monocular image sequence regardless of complex and changing background.We start from a mathematical viewpoint statistical regularization based on Markov Random Field(MRF),and proceed by mapping the algorithm onto the cellular neural network(CNN).The CNN-MRF model based on spatial-temporal neighborhood system is given and the cost function is decided by the model.A simulated annealing is adopted to get the better result.The desirable feature of CNN is that the processors arranged in the two dimensional grid only have local connections,which lend themselves easily to VLSI implementations.Experimental results from a real monocular image sequence demonstrate the feasibility of the proposed approach.