Research And Application Of Image Denoising Based On Memristive Neural Network

With the rapid developments of imaging and other related disciplines,the digital image technology is widely used in almost all aspects of human production and life.With the progress of science and technology,the number of images that need to be processed is increasing and the requirements for their quality are constantly improving,which puts forward higher requirements for digital image processing technology.Therefore,the neuromorphic systems with parallel computing characteristics receives particular attention from the researchers.At present,the sizes of semiconductors and transistors cannot be further reduced,which restricts the research and development of the neuromorphic systems that closely related to the electronic circuit technology.Fortunately,the appearance of the nano-sized memristor,which can be used to simulate the synapses between neurons,has realized the construction of memristor neural network.As such,more functions of human brain can be realized,which therefore lays a foundation for memristor neural network to be used in image processing.The research content of this paper is to analyze the characteristics of memristors and establish the connections with neural networks.Then,design memristive pulse neural networks and memristive Chebyshev neural networks for image denoising.Finally,two new denoising methods are used to denoise the oil pipeline images.The specific research contents are as follows:First,based on the memristor theory,the HP memristor model,the threshold adaptive memristor model and the magnetron memristor model are introduced,and the theoretical research and MATLAB simulation experiments are also carried out.The analysis focuses on the voltage threshold adaptive memristor model and the magnetron memristor model with a non-zero initial state,which shows that the memristor can be used as the synapse of the neural networks and realize the characteristics of the biological spike-time-dependent plasticity(STDP).Second,combining the voltage threshold adaptive memristor with the pulse coupled neural network(PCNN)model,the memristive pulse coupled neural network(MPCNN)model is proposed.The connection strength of the neural network is determined by the total output of the memristor,which realize the connection strengths between neurons can be adaptively and dynamically changed with external input changes.Furthermore,the imagedenoising algorithm of the model is proposed.Then,the memristor Chebyshev neural network is designed based on the combination of the magnetron memristor synapse and the Chebyshev neural network,which is applied to denoise the images with Gaussian noise.The both of two designed algorithms are performed by MATLAB software,and the feasibility and superiority of the denoising algorithms are verified by two evaluation indexes,i.e.,the peak signal to noise ratio(PSNR)and the structural similarity(SSIM).Finally,the above two developed memristor neural network denoising algorithms are applied to the image denoising of oil field pipeline,and the simulation results are evaluated from the subjective and objective aspects,which proves that the methods proposed in this paper has good denoising performance.