基于马尔可夫模型优化的非经典感受野轮廓检测算法

摘要

We present a new algorithm based on isotropic inhibition effect of non-classical receptive field.First,the proposed algorithm introduced multilevel surround suppression into isotropic detection method.Secondly,an improved combination method based on set theory and the idea of edge growth control was proposed,which had solved the influence of the suppression parameter on contour detection.Finally,according to the theory of Markov random field,we established contour probability model,and got the output contour.The experimental results show that the new algorithm has high precision and is superior to the traditional method.%基于非经典感受野的同性抑制作用提出一种新的算法.该算法首先将多级周边抑制引入到各项同性检测模型中;其次根据集合理论,结合边缘生长过程控制思想提出一种改进的组合方法,解决了抑制参数取值对轮廓检测的影响;最后根据马尔可夫随机场理论建立轮廓概率模型,得到最终优化后的输出轮廓.实验结果分析表明,新算法精度高,明显优于传统方法.