基于上下文MRF模型的彩色街景图像分类

摘要

Using the contextual Markov Random Field model can transform the image classification problem into the minimization problem of the energy function. This method constructs the prior observation field model between MRF and the color image of the street and use the iterative conditional mode algorithm to get the minimum energy of the posterior label field. The comparison of this algorithm with FCM shows that it is more effective and efficient than the FCM algorithm.%利用上下文马尔可夫随机场（Markov Random Field ，MRF）模型，将图像分类问题转化为能量函数最小化（最优化）问题。该方法构建了MRF关于彩色街景图像的先验观测场模型,并利用迭代条件模式（Iterated Conditional Model ，ICM）算法获得后验标记场能量最小。通过和模糊C均值（Fuzzy C-means，FCM）算法实验对比表明，该方法不仅能有效分类，而且分类精度要远高于FCM。