Research On Neutron Image Denoising Method By Improved RPCA

Neutron radiography technology is a non-destructive testing technology for materials with the help of neutron penetration.The decay coefficient of neutron varies from material to material while neutron penetrates the object.Neutron imaging technology takes advantage of the above characteristics to explore the material composition and morphological structure of the object.The technology effectively complements the shortcomings of X-rays and gamma rays detection methods and has unique advantages and characteristics,which makes it play an important role in the fields of defense industry,medicine and new material detection.The additive Gaussian noise caused by dark current and interference of electronic instruments are unavoidable flaws for neutron imaging.In addition,although the imaging system is equipped with shielding protection devices,high-energy gamma photons,secondary scattered neutrons and other radiation particles will still collide the CCD chip in the process of neutron beam generation,transmission and subsequent digital imaging,which causing in the image contains prominent random white spot noise.All the above factors will result in the loss of information,low resolution and contrast and blurring of neutron image,which will seriously affect the subsequent processing.Therefore,the research on how to denoise and restore the neutron image can not only improve the image quality,but also reduce the requirement of the image quality on the precision of the imaging system.Robust principal component analysis(RPCA)is a method to recover a low-rank matrix by the correlation between the information of rows and columns of the matrix.The model can only suppress the high-intensity signal(high-frequency noise and the high-amplitude information of the image).However neutron images not only contain various types of noise,the grayscale range of the image is relatively large.Aiming at above problem,in order to remove the Gaussian noise and random white spots noise in the neutron image,protect the edge and texture information,this paper proposes a novel denoising method based on a‘detection-location-removal' iteration mechanism,which introduced an accelerate proximal gradient algorithm to optimize the improved robust principal component analysis(RPCA)and extract large size and high range information of image.The random white spot noise is identified and located by combination with the appropriate threshold selection strategy and median filtering technology.The detected noise is gradually removed through the iterative process.Finally,the neutron image with good denoising effect is obtained.Compared with other methods,the experimental results of simulation test and practical neutron image prove that the proposed denoising method can not only effectively remove therandom white spot noise in the neutron image,but also retain the edge and texture features of the neutron images.In particular,the quantitatively evaluations also illustrated the performance and effectiveness of the proposed method superior to the other denoising methods.Therefore,our proposed method was feasible for providing an improved image quality in neutron radiography NDT application.