基于Markov随机场的木材表面缺陷图像分割方法

Image Segmentation Method of Wood Surface Defects Based on Markov Random Field

针对木材表面缺陷的彩色图像,建立Markov随机场分割模型,采用EM法估计了模型参数,采用模拟退火算法(SA)和条件迭代算法(ICM)求解分割结果。试验表明,Markov随机场能对木材表面缺陷图像进行有效分割,但对死结图像存在欠分割现象。SA分割效果要优于ICM,但SA算法计算过程复杂,用时较长。

Markov random field theory is used to establish the wood surface defect color image segmentation model. EM algorithm is used to estimate the parameters in the model, optimize and solve the segmentation problem by Simulated Annealing algorithm（SA） and Iterated Conditional Mode algorithm （ICM）. The results showed that the segmentation method based on MRF can segment the wood surface defect images efficiently and usefully, but there was deficient in segmentation of hard knot images. The segmentation result of SA was better then ICM, but SA segmentation methed was complicated, and consume more time.