Application Research On Multi-scale Image Fusion Based On PCNN

Pulse Coupled Neural Network(PCNN),regarded as a new model of the third generation artificial neural network,has been applied to various fields of image processing.Its biological background is to simulate the visual cortex of cats and other animals,it makes use of the two characteristics of the neuron: linear addition and nonlinear multiplication modulation coupling,and considering the characteristics of the transmission delay and the exponential decay,the synchronization of the adjacent neurons of the visual system of the animal,and the balance of the internal activities when the neurons are in the inhibitory state,so that this model can be more in line with the actual biological neural network.Pulse coupled neural network model is a single layer neural network,which can be used for feature extraction,image segmentation,image fusion,pattern recognition and so on,so it is very suitable for digital image processing.This thesis focuses on the application and research of pulse coupled neural network in the field of image fusion.The pulse coupled neural network is widely used in image fusion,but there are many parameters need to be set up by artificial experience in the model.Parameter setting is very important,it directly affects the fusion results.In this this thesis,the method of setting the parameters of PCNN model is studied,a new algorithm based on genetic algorithm optimized PCNN model combined with the nonsubsampled contourlet transform in image fusion is proposed.This method can be used to set the parameters of the model adaptively,which reduces the number of parameters needed to be set by artificial experience,and avoids the blindness in the process of parameter selection.The feasibility of the algorithm is verified by simulation experiments.After that,the PCNN model is improved,the improved pulse coupled neural network model has replaced the gray-scale value of the image and introduced the weighted product of the strength of the image's gradient and the local phase coherence as the model input,which makes the image fusion effect of pulse coupled neural network and nonsubsampled contourlet transform better.In view of the successful application of PCNN in the nonsubsampled contourlet transform of multi-scale image decomposition tools,this thesis introduces the network model into the bidimensional empirical mode decomposition,a new method of multi-scale image fusion based on the combination of pulse coupled neural network and image compression is proposed.At first,BEMD processes the original images decomposed into multiple bidimensional Intrinsic Mode Function(BIMFs)and a residual image.Then after doing compression measurement on each layer of BIMFS,we can compression measurement coefficients.The coefficients at the same layer do the PCNN image fusion and we can get measurement sampling BIMFs.And then after measurement sampling BIMFs reconstructed,we can get the final BIMFs.The residual images do the fusion based on entropy weight to get the final residual image.At last,the final BIMFs and the final residual image,after BEMD inverse transform,to get the result image.The simulation results show that the fused image of PCNN and bidimensional empirical mode decomposition also has a good fusion effect.