Research On Fast Neutron Image Denoising Algorithm Based On Total Variation Model

Fast neutron radiography as an important supplement of existing non-destructive testing technology,in the aerospace and counter-terrorism,etc have important potential application advantages,but due to the fast neutron extremely difficult to detect,lead to the signal-to-noise ratio of the Fast neutron radiography image is extremely low,on the useful information in image read or discrimination,must first uneffective noise reduction processing.In practice,some conventional noise reduction algorithms are more or less limited and cannot meet the noise reduction requirements of fast neutron images.Based on the study of the total variation denoise model,this paper improves the total variation denoise model according to the characteristics of fast neutron image noise.The main work is as follows:(1)according to the principle and device composition of fast neutron photography,the main noise sources in fast neutron photography and the main noise characteristics in fast neutron image are analyzed.(2)according to the characteristics of fast image noise density is high,in the traditional total variation denoise model introduced the local gradient operator,make full use of the self-similarity of image,improve the total variational denoise model easy to some larger size when dealing with high density noise as the noise of the image detail to keep the disadvantages,the experimental results show that the new improved algorithm can better fast reunion in the images of the shape to filter out noise.(3)the image quality is poorer,due to the fast image resolution is low,in order to avoid the total variation denoise model of further loss of image resolution after noise reduction processing,the introduction of neighborhood difference of statistics to pixels fuzzy features,its main idea is that the noise pixels,as far as possible processing was not without processing noise pixels,while the adaptive determination of total variational noise parameters in the model,optimize the regularization and the relationship between the fidelity term.(4)using Visual Studio2012 wrote a fast image processing software,based on the above algorithm and fast image quality of noise reduction after processing has carried on the analysis of subjective evaluation and objective evaluation of the experimental results show that the new improved algorithm in comparison with the conventional noise reduction algorithm can more effectively filter out noise fast neutron images,have excellent image detail and edge protection at the same time.