Research On Image Segmentation Algorithm Based On PCNN Model

In recent years,with the accelerated process o f information technology and the rapid development of computer technology,people in life and production of computer vision needs more and more urgent.Computer vision is the use of computer simulation of biological vision system,the perception and understanding of the environment,it is in the military,industrial production and intelligent monitoring and so has great potential value.Image segmentation is the first step in computer vision system for image processing,one of the core technologies of compu ter vision.At present,the commonly used image segmentation algorithm has threshold method,edge detection method,active contour model segmentation method and so on.Based on the study of classical segmentation algorithm,this paper proposes an Improved Pulse Coupled Neural Network(IPCNN)image segmentation method for segmentation of images in complex background environment.The main research work is as follows(1)In view of the lack of uniform evaluation criteria for the current image segmentation algorithm,this paper has carried on the detailed analysis to the existing performance appraisal criterion of the image segmentation algorithm,has laid the theoretical foundation for improving and put forward the new evaluation criterion.At the same time,considering that most of the original images are polluted by noise,pre-processing is needed before segmentation.Therefore,several kinds of classical image preprocessing methods are analyzed and experimentally studied to test their processing effect,for the subsequent image segmentation provides a guarantee.(2)The principle of classical algorithm such as threshold segmentation method,Snakes model segmentation method and C-V model segmentation method is summarized,and the problems existing in the above algorithm and its applicable environment are analyzed and compared,which lays the foundation for the research of image segmentation in the complex background environment.(3)The mathematical model and working principle of PCNN are discussed in detail.In the complex background environment,the traditional image segmentation algorithm has the advantages of low segmentation precision and poor anti-jamming.In this paper,the image segmentation algorithm based on PCNN model is proposed by simplifying the coupling characteristics of PCNN model neurons.On this basis,the correlation between the pixels is considered,and the PCNN model is improved by introducing the multi-threshold idea.Experiments show that the segmentation effect,anti-jamming,computing speed and stability of the algorithm are larger.(4)Aiming at the problem that the parameters of PCNN model are complex and varied,and the selection of parameters has great influence on segmentation effect and efficiency.In this paper,we propose a two-dimensional maximum interclass variance method to optimize the initial threshold.In order to improve the real-time performance of the algorithm,the related fast recursion formula is deduced and given;at the same time,different from the traditional PCNN based on experience or through a large number of experiments to determine the key parameters of the practice,but from the coupling characteristics of PCNN,combined with the image of their own space and gray features,by calculating the local gray squared variance to determine the connection strength coefficient,Finally,according to the maximum principle of information entropy to determine the segmentation results,to achieve the target object adaptive automatic segmentation.In this paper,we consider the difference between the spatial and gray values of the pixel.The numerical experiments show that the proposed algorithm has the advantages of fast image segmentation,clear target segmentation and strong anti-jamming effect compared with the traditional PCNN algorithm.