Image Fusion And Enhancement Based On Pulse Coupled Neural Network And Multi-feature Analysis

As the important technologies to improve the utilization rate of image information and improve image quality,multi-source image fusion and enhancement have wide application foreground in many fields such as medical diagnosis,military target recognition and satellite remote sensing image processing.At present,most fusion and enhancement algorithms are supervised and optimized based on image quality evaluation function,which can obtain better objective evaluation results.But there are some problems of them,such as lack of precision,poor visual effect,and high complexity.To address these problems,this thesis combines pulse coupled neural network(PCNN)with multi-feature analysis to develop multi-source image fusion and enhancement algorithms.The main research contents are as follows:Firstly,the characteristics of PCNN's neuron link,the pulse release and time signal dynamic behavior are analyzed.The advantages of PCNN model in image processing are verified by the experiments.In addition,we analyze the image feature correlation information from three aspects: the image pixel clustering characteristics,image quality evaluation function and related statistical characteristics,image subjective visual perception characteristics,which lays a theoretical foundation for proposing relevant algorithms.Secondly,aiming at the problems of the declined infrared target and lack of contrast of visible light in the fusion of infrared and visible images,we proposed a new infrared and visible image fusion algorithm based on PCNN and target extraction.The Salient Object Detection(SOD)algorithm based on image pixel clustering is used to segment the infrared target information,then the segment results are fused in Nonsubsampled Contourlet Transform(NSCT).The experimental results show that the proposed algorithm can highlight the significance of infrared targets and retain the details of visible images,which is superior in visual observation.Thirdly,a multi-focus image fusion algorithm based on PCNN combined with visual statistics is proposed according to the analysis of image quality evaluation function and image visual statistical characteristics.By analyzing the statistical characteristics of the image,the multi-focus image area is divided into focus,non-focus and boundary regions,and the regional definition evaluation function and regional fusion criterion are established.Experimental results on multi-focus datasets show that the proposed algorithm effectively overcomes the problems caused by noise,blur and misregistration in pixel fusion methods,and it has faster computational speed than convolutional neural networks.Finally,aiming at the difficulties faced by the current algorithm in processing the original image color distortion,noise and low contrast,an image enhancement framework based on PCNN and Human Visual Perception(HVP)is proposed and applied in underwater image enhancement.By analyzing the color composition of color images,color space,and the distribution of natural images in color space,the human visual preference model is obtained.Based on the human visual preference,the image to be enhanced is color-improved and the image details are then enhanced by the PCNN model.The experimental results show the feasibility and effectiveness of the algorithm,which can effectively solve the problem of poor visual observation caused by color distortion.