Research On Image Segmentation Based On PCNN

At present,Pulse Coupled Neural Network(PCNN)is widely used in the research of image segmentation methods.The model parameters are too complicated,the segmentation time is too long,and the parameters need to be adjusted manually.The model itself is not suitable for noise images and the traditional PCNN.The single-channel ignition feature does not allow accurate positioning of the target pixel.In this context,this paper will study the PCNN parameter adaptability,improve the image noise immunity,improve the model optimization criterion and multi-channel fusion.The main work of this thesis is as follows:1.In this paper,based on a simplified PCNN model,the image direction information measurement method is used to realize the adaptability of the coupling strength coefficient.2.In this paper,the Laplacian operator and Gaussian function are used to design the matrix of the connection coefficients of the feedback input domain,so that the image can effectively protect the edge details during the segmentation process and also has anti-noise.3.In this paper,a double-threshold judgment criterion is proposed.This method introduces the maxinum inter-class variance method for initial ignition and generates suppression neurons,which suppresses the ignition of non-target pixels based on the model of ignition discrimination.4.This paper proposes an improved PCNN image segmentation method combined with grayscale weight adjustment.The method uses the OTSU algorithm to measure the multi-channel targets that have been segmented,and combines them according to the measurement information.