CIS Signal Dependent Noise Cancellation Method Based On Stochastic Resonance Dynamic Mechanism Analysis

CMOS image sensor(Complementary Metal-Oxide Semiconductor Image Sensor,CIS)is the core component of image generation,which is widely used in image and video fields.However,in the process of image generation,it will be interfered by the signal-dependent noise generated by the CIS,and the image quality will be affected.In order to improve the signal-to-noise ratio of the output image,it is necessary to suppress and eliminate the CMOS image sensor signal dependent noise.This is a key step in the imaging steps of CMOS image sensors,so it is of great significance for its analysis and research.This thesis summarizes the current research status of CMOS image sensor signal-dependent noise suppression processing at home and abroad.It is found that noise suppression algorithms of existing CMOS image sensors consider noise to be harmful.In order to remove noise,the original details of the image are lost.The noise suppression method based on stochastic resonance can use noise to enhance the original image signal,but the existing image noise suppression algorithms based on stochastic resonance are all aimed at additive Gaussian noise,and there is no stochastic resonance noise reduction algorithm based on signal dependent noise.Aiming at the above problems,this paper designs a local piecewise linear noise level function to estimate the noise parameters for the CMOS image sensor signal dependent noise model,and then designs a signal dependent noise elimination method based on the analysis of the stochastic resonance dynamics mechanism.The main steps are as follows:Different from the global noise level function of the existing papers,this paper divides the noisy image into blocks,calculates the noise variance and gray intensity combination of the local image block,and forms the local segmented noise level function,which is estimated The noise parameter of the partial image block.Different from the existing analysis and research of stochastic resonance image noise reduction algorithms based on adiabatic approximation theory,this article is the first time based on the theoretical system of nonlinear dynamics to study and analyze the stochastic resonance system around attractor theory.Analyze the transition process offset of stochastic resonance to study the change rule of stochastic resonance transition threshold and transition width parameter after the offset change in the transition process.According to the estimated local noise parameters,the transition threshold and transition width parameters for the optimal signal-to-noise ratio of the image are calculated,and stochastic resonance processing is performed according to the obtained parameters,thereby effectively eliminating the CIS signal-dependent noise.Based on the signal-dependent noise model,this thesis adds 30 kinds of noise parameters to 25 noise-free images to form a test noise image.This thesis uses these test images to verify the proposed signal-dependent noise cancellation method based on stochastic resonance dynamic mechanism analysis.Experimental results show that,compared with the existing noise reduction algorithm,the proposed algorithm can increase the peak signal-to-noise ratio(PSNR)value by 15.11%under the condition of noise parameters(k1>0.2,k0>0.2),and the structural similarity(SSIM)value increased by 37.27%.