LOW-DOSE CT IMAGE RECONSTRUCTION BASED ON WAVELET-TGV AND NON-CONVEX PENALTY FUNCTION

Reducing the computed tomography (CT) scanning dose decreases the risk of cancer; however, low-dose imaging is accompanied by a high level of noise. The present paper focuses on the noise statistical characteristics of low-dose CT projection data and establishes a low-dose CT projection domain denoising based on wavelet-total generalized variation (wavelet-TGV) and non-convex penalty function with a view to promoting the quality of reconstructed images. The model utilizes the high-order derivative information of the total generalized variation, which can avoid the staircase effect and block artifacts introduced by the TV minimizing denoising process, further enhancing the sparseness of the objective function under the constraints of the non-convex penalty function. The Split Augmented Lagrange Shrinkage algorithm was exploited to optimize the solution, and the reconstructed results of the proposed denoising method were experimentally compared with those of other methods. In terms of visual effects and quantitative analysis, we demonstrate that the quality of the image reconstructed using the present denoising method can be significantly improved while preserving the boundary and textural details.