Infrared Image Segmentation using Hidden Markov Random Fields and Expectation-maximization Algorithm

摘要

Circuit board infrared image segmentation is an important procedure in the application of circuit board fault detection with infrared thermal imaging technology. A CNO-HMRF-EM algorithm combined with the advantage of HMRF, EM and CNO is designed to deal with the insufficiency of the traditional clustering methods in the circuit board infrared image segmentation . To get the best clustering segmentation results, HMRF-EM algorithm is used as the first step to estimate the tag of each image point so that each point of image can be clustered according to the tag estimation result. Then the HMRF-EM algorithm?s optimal clustering number is determined in the use of CNO algorithm. The simulation results prove that, comparing with the methods of C-Mean clustering and OTSU clustering, bigger GS value as well as the better results of the clustering segmentation can be acquired in the use of CNO-HMRF-EM algorithm.