Elimination Of Fixed-mode Noise In Columns In High-resolution Solar Images

CMOS image sensor has the merits as low power,small size and high level of integration,which provides great support for high resolution solar image acquisition.And,the acquisition of high-resolution solar images provides strong support for scientific research,such as solar atmosphere and solar physics.However,the mismatch between processing circuit and A/D converter of the detectors and the difference between the sensors can cause column-fixed pattern noise(CFPN)in the image data.The presence of these noises not only reduces the quality of the image,obscures the details of the image,but also influences the scientific research that revolves around these images.Therefore,it is necessary to eliminate the CFPN.In recent years,many scholars at home and abroad proposed a variety of algorithms based on the distribution,statistical laws of CFPN.Such as,statistics,variational,Fourier filtering and multi-scale de-noising method.However,due to the limitations of these algorithms,the image after de-noising becomes blurred and distorted.In order to solve the above problems,a robust de-noising algorithm was proposed,which is based on the wavelet transformation and“dual filters",in this thesis.Firstly,the implementation of the algorithm was demonstrated by simulation experiment.The main execution of the algorithm is:According to the formation mechanism,noise characteristic and the behavior in the original image of the CFPN,the logarithmic operation is carrying out before decompose the noisy image.Through statistical analysis,the noise-free and noise wavelet coefficients of vertical component in the wavelet domain are modeled as Gaussian distribution.Thus,the median filter is applied to wipe off the noise wavelet coefficients.Next,using the inverse wavelet transform to obtain the noise-free image,and subtracting it from the logarithmic image to extract initial CFPN.However,the low frequency domain of initial CFPN contains some noise-free signal.So,the low-pass Gaussian filter is applied to get the final CFPN.Finally,using the original image divided by final CFPN to get the de-noised image.In order to verify the accuracy and validity of the proposed algorithm several comparative experiments with existing de-noising methods are conducted.The simulation results show that the proposed algorithm can eliminate 94 percent of CFPN.The mean value,peak signal-to-noise ratio(PSNR),structural similarity(SSIM)and power spectral density shows that the proposed algorithm can get better de-noising result.Furthermore,In order to inspect the effect of thresholds on the results,several different thresholds were test.The results show that the proposed algorithm has a strong reflection on the width of the median filter window and the Gaussian kernel.Two high-resolution image sequences are processed by the algorithm,Which taken from the New Vacuum Solar Telescope(NVST)at the Fuxian Solar Observatory(FSO)and New Solar Telescope(NST)at Big Bear Solar Observatory(BBSO),respectively.The results show that the algorithm can not only remove the noise accurately,but also can keep more image details,and the image feature is more obvious.At the same time,the results of each evaluation index are obtained by this algorithm,is also the best.