基于二维隐Markov模型的图像分割新算法

New Image Segmentation Algorithm Based on 2D Hidden Markov Model

提出了一种新的二维隐Markov模型解码准则,利用贪心法求得次最优解,在一些假设的基础上推导出计算上可行的递归形式;并将新的解码准则应用于图像分割,实现了一种迭代的图像分割算法。每一次迭代中,解码过程按照自上而下、自左向右的顺序进行。此外,还针对基于二维隐M arkov模型的图像分割方法存在的常见问题,给出了简单有效的后处理方案,消除了孤立点,取得了较好的分割效果。

XA new decoding criterion for 2D hidden Markov model is proposed and a local optimal solution is provided with a greedy algorithm. For computational feasibility, the recursion is derived based on some assumptions. The new criterion is applied to image segmentation and an iterative image segmentation algorithm is implemented. In each iteration, the decoding is done from top to bottom and from left to right. Furthermore, an easy and effective post-pro- cessing method is adopted to overcome the common problem in image segmentation based on 2D hidden Markov model. The method eliminates isolated points and obtains good segment resuits.