Research Of Image Fusion Based On Nonsubsampled Contourlet Transform And Pulse Coupled Neural Networks

Recent years,image fusion has been widely used in both military and civil fields.Image fusion aims to study how to extract the features and integrate the information of multiple source images and get an image with better quality and richer information.The pixel level fusion algorithm based on transform domain is a hot topic at present,it can get good fusion effect,and its elements are multi-scale analysis tools and fusion rules.Nonsubsampled contourlet transform(NSCT)has many advantages,such as directivity,anisotropy and translation invariance,which is more suitable for image fusion than other multi-scale analysis tools.Pulse coupled neural networks(PCNN)is a feedback neural network model,which has the characteristics of global coupling and impulsive synchronization.Using PCNN as a fusion rule in NSCT,we can get high-quality fusion images.However,when the traditional PCNN is applied to image fusion,the gray value of a single pixel is used as external excitation directly,which cannot preserve the details of the image very well,and leads to lower contrast and details loss easily.Therefore,a new method of image fusion is proposed in this paper,which combines NSCT with PCNN.For fusing the subbands decomposed by NSCT,the modified spatial frequency is used to motivate the simplified PCNN model,and the coefficients with large firing times are selected.The fusion experiments of multiple groups of different test images are carried out,and we have compared the results with the related fusion methods.It is proved that the proposed fusion method has better fusion effect and stability.Finally,the fusion of infrared multifocus images and infrared-visible images combined with proposed method is realized respectively.