X-ray Image Denoising Using Blind Source Separation In Anscombe Domain

In the process of digitization or transmission of X-ray images,noise is always affected by equipment and environmental factors.Eliminating noise has always been an important research topic in the field of X-ray imaging.Blind source separation(BSS)is an excellent signal processing method.By analyzing the correlation of observation data,the hidden independent signal components are found under the condition of unknown mixed channel parameters to complete the blind separation of source signals.This paper focuses on the problem of X-ray image denoising,and uses high-order nonlinear principal component analysis(NLPCA)and second-order weight-adaptive variant second-order blind identification(WASOBI)statistical methods in Blind Source Separation(BSS)to blindly separate the signal and noise of the X-ray image sequence.The main research work is as follows:(1)The basic concepts and principles of Blind Source Separation(BSS)are introduced,and analyzes the feasibility of Blind Source Separation(BSS)for image denoising;An algebra-based blind source separation method is studied.The blind source separation(BSS)based on algebra is studied.The optimization function and initial value of the algorithm are determined by analyzing the hidden data structure information in the observation data to achieve blind separation.The principle and generalized form of Anscombe transform are introduced,and the conversion of non-Gaussian noise types is realized.(2)For multiple noisy X-ray images,a method of denoising based on nonlinear principal component analysis NLPCA based on higher-order statistics is used.NLPCA is based on the original principal component analysis(PCA)algorithm,implicitly introduces higher-order cumulants,and optimizes the objective function using a nonlinear function.The denoising method samples a sequence of X-ray images and converts Poisson noise into Gaussian noise in the image through Anscombe transform;Then consider each noise-containing image as a combination of noise and signal components,and then use NLPCA to separate the signal and noise components to achieve the purpose of denoising;Finally,the final denoising image is obtained by inverse transform of Anscombe.Experimental results show that this method can effectively remove X-ray image noise,and the denoising effect is significantly better than traditional denoising algorithms.(3)For the blind source separation(BSS)problem where the observed signal has a time-series structure,only the second-order statistics are sufficient to estimate the mixing matrix and sourcesignal during the separation process,so the weight-adaptive variant second-order blind identification(WASOBI)algorithm based on second-order statistics is used to replace the NLPCA in the research process of(2)above.The WASOBI algorithm sets an adjustable weight matrix to minimize the mean square error of the mixed signal,so the estimated source signal is closer to the actual signal.The experimental results show that: WASOBI can accurately adjust the appropriate weight parameters during the separation process,improve the accuracy of the algorithm separation,and the running time is short.Optimize the process of separating image signals from X-ray image sequences based on blind source separation(BSS).