Abstract: Objective This paper proposes a novel CT reconstruction model by leveraging the sparsityinducing property of L1/L2 regularization to enhance image quality. Methods A weighted matrix is incorporated into the L1/L2 total variation functional space to improve structural feature representation. To address the nonconvex and nonsmooth optimization challenges inherent in the model, an augmented Lagrangian framework with auxiliary variables is introduced, which enables decomposition into tractable subproblems that are iteratively solved via the alternating direction method. Conclusions Compared with conventional regularization approaches, the proposed weighted L1/L2 model demonstrates superior performance in limited-angle projection image reconstruction, particularly under severe Gaussian noise conditions, achieving enhanced reconstruction accuracy and artifact suppression.