结合博弈论的马尔可夫随机场红外图像分割  

Infrared Image Segmentation Algorithm Based on MRF Combined with the Game-theory

作　　者：王坤[1] 张恺[1] 王力[1] 诸葛晶昌[1] 

机构地区：[1]中国民航大学,天津300300

出　　处：《红外技术》2014年第10期801-806,共6页Infrared Technology

基　　金：中国民航大学校级项目基金;编号:2010kyE07;2010QD02S;国家自然科学基金;编号:U1333111;天津市自然基金;编号:12JCQNJC00600;中国民航机务维修科研基地资助

摘　　要：电路板红外图像芯片提取是电路板红外图像故障检测系统中的重要环节,已成为红外图像分割领域关注的一个重点。传统的芯片发热区域提取方法大多需要人工干预,且分割效果不理想,容易丢失边缘信息,导致细节特征不明显。针对以上缺陷提出一种结合博弈论的改进马尔可夫随机场分割算法。首先用现有的OTSU算法对图像进行粗分割,将图像分为两个子集(背景域和目标域),然后利用马尔可夫随机场(MRF,Markov Random Field)理论建立图像分割模型,最后利用结合博弈理论的MMD(Modified Metropolis Dynamics)算法,根据模型分别对每个子集进行细致分割,提取核心发热区域。实验表明,改进算法应用在电路板芯片发热区域提取时,能够较好地抑制噪声,准确处理边缘信息,与传统算法相比,在视觉效果和客观数据上都有很大的提高,具有一定的准确性和鲁棒性。Extracting the infrared image chip exothermic area at a circuit board,which is an important part of the fault detection system, has also been a key point in field of image segmentation. Traditional extraction methods need manual intervention more or less, and the results aren＇t ideal, also making the edge information and detail features be lost easily or not obvious. In view of above shortcomings, a segmentation algorithm based on improved Markov random field combined with the game theory is proposed. Firstly, use the OTSU algorithm for crude segmentation and divide the image into two subsets（background and target domains）; Secondly, establish the image segmentation model through MRF; Thirdly, segment each subset meticulously and extract exothermic areas through MMD（Modified Metropolis Dynamics） algorithm combined with the game theory according to the model. The experiment results show that when extracting the infrared image chip exothermic area, the proposed algorithm can not only make the speckle noise suppressed, but also keep the detail characteristics effectively. Compared with the traditional ones, it makes a great improvement on the visual effect and objective data, and has a certain accuracy and robustness.

关 键 词：红外图像分割 马尔可夫随机场 博弈论 OTSU法 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]