Applications of evolutionary programming in markov random field to IR image segmentation

A new image segmentation algorithm based on Markov Random Field (MRF) with Evolutionary Programming (EP) is presented in this paper. As Infrared (IR) image has blurry edges and fuzzy texture, the segmentation of infrared image becomes more complicated and sophisticated. MRF model is an effective way for segmenting fuzzy image, which has been used in many field of image processing. However the segmentation algorithm based on MRF must optimize the image field, and the classical optimization algorithm is Simulated Annealing (SA) that could get the global optimal resolution with heavy calculation burden. To avoid the unacceptable calculations, we use Evolutionary Programming (EP) algorithm to describe the optimizing process of MRF model. Evolutionary Programming is a heuristic algorithm that emphasizes the evolution of individuals in a neighborhood instead of a pixel. It permits EP algorithm to access the global optimization faster than SA algorithm. Furthermore the 'Survival of the fittest' ideas are introduced into MRF model, which could describe the correlations of pixels or individuals in a neighborhood. The coactions and competitions could strongly constrain the noise and blur edges. The followed image segmentations experiments proved that this novel algorithm could reach a satisfied result among the noise restraint, edges preservation and computation complexity.