Research On Pulse Coupled Neural Network And Its Application In Fusion Of Infrared And Visible Images

With the continuous development of computer vision,image processing technology is more and more applied to actual scenes.The fusion of infrared and visible light images has a broad application prospect as a research hotspot in image processing.The fusion of infrared and visible images by pulse coupled neural network(PCNN)is a hot and difficult issue in this field.This subject uses multi-scale analysis as a tool,and uses the pulse coupled neural network(PCNN)theory to study the accuracy and reliability of infrared and visible image fusion.Firstly,based on the pulse-coupled neural network,a binary function is used to simulate the output characteristics between neurons based on the traditional PCNN model structure,which leads to the problem of losing a lot of local correlations between neurons,an S-type dynamic output coupled pulse-coupled neural network(SPCNN)is proposed.The proposed model strengthens the strong and weak relationship between neighboring neurons,which is beneficial to the acquisition of image detail information.The proposed SPCNN model is combined with nonsubsampled contourlet transform(NSCT)to lay the foundation for reliable infrared and visible image fusion.Secondly,in the traditional PCNN model,although the threshold value changes continuously with the iteration,this change is a fixed value,which makes the threshold value lack of adaptability for different image inputs.So based on SPCNN,a threshold adaptive pulse coupled neural network(TAPCNN)is proposed.The proposed TAPCNN constructs the activation difference function,adaptive attenuation coefficient and adaptive threshold function,so that the TAPCNN enables the network model to adaptively adjust the threshold according to different pixel value distributions of the input image,and provide a guarantee for obtaining rich texture and background information for the fusion of infrared and visible images.Thirdly,based on the theory of the fusion strategy of infrared and visible images,an improved image fusion method of secondary decision is proposed.The secondary fusion of the source visible image and the primary fusion image is helpful to improve the clarity of the secondary fusion image.In addition,according to the relevant theoretical basis of the high-frequency sub-band image,the high-frequency fusion rule is improved.The improved rule makes the information such as texture and edge in the high-frequency sub-band image more prominent..Finally,by combining the proposed TAPCNN with nonsubsampled shearlet transform(NSST)to achieve more accurate infrared and visible image secondary decision fusion,and the accuracy and reliability of the improved algorithm were verified through experiments of infrared and visible image fusion.