Research On Image Denoising Method And Application Based On Composite Multistable Stochastic Resonance

Stochastic resonance image denoising is an important research direction in digital signal processing.The classical bistable stochastic resonance model suffers from high potential barriers and easy saturation,while the model parameters are mainly adjusted manually or by qualitative analysis,which is time consuming and cannot achieve optimal resonance.To solve the above problems,this paper studies the following contents:1.Aiming at the problems of high barrier,easy saturation and low system stability of existing stochastic resonance models,a composite multi-stable stochastic resonance model is proposed.Firstly,a novel type of stochastic resonance potential well model is analyzed.The model solves the shortcomings of high barrier and easy saturation,but the system stability is poor and the problem of deep middle potential well is prominent,which affects the signal enhancement ability of stochastic resonance.In order to solve this problem,the Gaussian model is integrated into the novel type of potential well model to construct a composite multi-stable stochastic resonance model.The theoretical feasibility of this model is verified by Kramers escape rate and output signal-to-noisc ratio.2.Aiming at the difficulty of model parameter adjustment,an adaptive composite multistable stochastic resonance image noise reduction system is constructed.This paper analyzes the disadvantages of using full-reference evaluation index of adaptive stochastic resonance at present,introduces the block-noise standard deviation of no-reference evaluation index as the fitness function of adaptive algorithm,and selects whale algorithm for adaptive optimization of model parameters.Then,different images containing Gaussian,salt and pepper noise and multiplicative noise are denoised through simulation experiment.The effectiveness of complex multi-stable stochastic resonance noise reduction system is verified,and the feasibility of blocknoise standard deviation is also verified.3.The composite multistable random resonance array system is investigated and applied to noisy image processing to address the problem that different scan sequences of the same image are processed by random resonance.The characteristics of noisy radar images and fog images are analysed,and the composite multistable random resonance array system fused with the total set averaging method and the cascaded composite multi-stable random resonance system are applied to radar image noise reduction and fog image noise reduction respectively.The experimental results confirm that the composite multi-stable random resonance array system with different array structures can improve the noise reduction effect of the noisy images to a certain extent.