Research And Implement Of Image Segmentation Based On Pulse Coupled Neural Network

Digital image processing is generated in the1950s, from then on the academiahave been studying it endlessly. A decade later, the digital image processing is indepen-dent out into a course.Image segmentation technology is an infrastructure to processimages, accounts for an important position in the field of image processing, and playsan important role in terms of image engineering, pattern recognition, computer vision.The research topic of this article just is image segmentation, the research purpose of it ishow to segment images time-saving and effectively, this paper proposes a method basedon the pulse coupled neural network of maximum entropy. So the research for segmen-tation method in this paper is certainly valuable.Traditional processing method of image segmentation contains these ways as be-low: basing on a grayscale arithmetic mean, basing on the entropy and the histogram,basing on the method of Otsu segmentation, basing on edge detection, basing on thre-shold segmentation, basing on boundary extraction segmentation, basing on regionsegmentation and basing on a particular theory. At the same time, it is necessary toemerge some related model, for example robust scale regional model，Two-dimensionalclassical C-V model, scale regional fitted model, three-dimensional C-V model, robuststatistical3D C-V model, geometric active contour model，and so on.In the paper, the method based on pulse coupled neural network (PCNN) will beproposed. This method can compensate some shortages above-mentioned when digitalimages are divided. In a digital image segmentation model of PCNN, a neuron is thebasic unit, it constitutes a two-dimensional single-layer neurons array. The number ofneurons in PCNN model is consistent with the number of pixels. Each neuron and eachpixel is a one-to-one correspondence. According to pulse propagation characteristics ofPCNN which cause the phenomenon of sync pulse，image segmentation is implemented.Although PCNN method has its indelible advantage. But there are still some dis-advantages. So the PCNN will be made better in this article, this improved PCNN useslinear manner to adjust the threshold value dynamically, uses maximum entropy to de- termine PCNN loop iterations, and uses the thinking of median filter to improve the re-ceiving part of the PCNN, so that to overcome the impact of noise on the segmentationprocess. Finally, the algorithm is applied to image segmentation. It is evident throughconsequences got by some experiments this paper’s method can be well adapted to dig-ital image segmentation, has strong universality, has short split time and well segmenta-tion effect.