The total generalized variation (TGV) method effectively reduces noise and mitigates the staircase effect commonly observed in total variation (TV) approaches . However, its capacity to preserve structural features and fine details within images remains limited, making it less effective in certain applications . To address these shortcomings, the paper introduce overlapping group sparsity (OGS) regularization into the first-order gradient of the TGV model. This integration significantly enhances denoising performance while further minimizing staircase artifacts. By leveraging the sparse structural information embedded within images, our method achieves superior noise suppression compared to conventional TGV models. Furthermore, we replace the commonly used L2 norm in the data fidelity term with the L1 norm, which better preserves undistorted image regions, particularly in cases of complex noise. To efficiently solve the resulting optimization problem, we employ a split Bregman method, ensuring computational efficiency. Experimental evaluations highlight the superiority of our approach, showing that it consistently outperforms state-of-the-art TV- and TGV-based methods in both noise reduction and detail preservation, delivering significant advancements in image denoising techniques.