The Application Study Of BM3D Frames And Variance Stabilizing Transformation On Neutron Radiographic Image Restoration

Neutron radiography is a non-destructive testing technique technology which displays internal structure and details of objects using neutron penetration.Compared with other detection techniques,it has unparalleled advantages,especially in terms of testing materials containing hydrogen,boron,lithium and other light elements are more prominent.Therefore,neutron radiography has an extensive applications in the fields of aerospace,industrial engineering,military science,medical imaging,etc.In the process of imaging,due to effects of the physical conditions and neutron scattering,γ-ray imaging system pollution and other factors,it will lead to serious degradation of neutron image,such as noise pollution,low contrast,image blurring,etc.It brings interference on testing substance and identifying the details of the internal structure.Thus recovering neutron image by denoising and debluring process is extremely important.In addition,it can reduce front-end imaging systems for neutron imaging hardware equipment requirements and costs.Block matching three-dimensional filtering(BM3D)is a non-local,multi-scale filtering technique utilizing image structure information.BM3 D frames based of variational image deblurring shows strong performance and robust quality in the image restoration under Gaussian noise.However,the actual neutron image noise type complex,for this problem,variance stabilizing transform is introduced to convert mixed noises to Gaussian noise approximately.Then the transformed data is filtered by BM3 D frames adaptively,finally restoration is complete by the variance stabilization inverse transform.The experiments on simulated images and neutron images can illustrate that the proposed restoration scheme based BM3 D frames and variance stabilizing transformation improves the image quality and possesses a robust performance.Especially in restoring the neutron image which is degraded severely,the proposed scheme retains the details information under the premise of suppressing noise effectively.