Neutron image denoising method based on adaptive new wavelet threshold function

Abstract

Neutron radiography is an important nondestructive testing technique in the industrial field. Due to the influence of various factors in the neutron imaging process, the image quality will be reduced. Aiming at the Gaussian-Poisson mixed noise in neutron images, an adaptive new wavelet threshold function denoising method is proposed to reduce the influence of noise on neutron images. The basic idea is to transform the Gaussian-Poisson mixed noise into a noise that approximately follows the Gaussian distribution by generalized Anscombe transformation, and then use Particle Swarm Optimization (PSO) algorithm and wavelet threshold function denoising method to remove the mixed noise in the image. The new wavelet threshold function overcomes the problems of discontinuity in the traditional hard threshold function and fixed deviation in the wavelet coefficient of the traditional soft threshold function, and has the adjustment factor, which can combine the advantages of the traditional soft and hard threshold functions. PSO algorithm is used to find the optimal adjustment factor for image denoising. In addition, the new wavelet threshold function is continuous and smooth at the threshold, avoiding excessive strangulation of wavelet coefficients. By analyzing the fixed threshold defect, a threshold selection formula that decreases with the number of decomposition layers is proposed. The results of Matlab software experiments show that the new method can significantly improve the Peak Signal-to-Noise Ratio (PSNR) and reduce the Mean Square Error (MSE) of noisy images compared with the other four methods in removing Gaussian-Poisson mixed noise. Therefore, the new method can retain more image details and effectively improve the quality of neutron images.