Core Concepts
The proposed weighted structure tensor total variation (WSTV) model effectively captures local image features and maintains details during the denoising process, outperforming other TV-based methods.
Abstract
The content discusses a new image denoising model based on weighted structure tensor total variation (WSTV). The key points are:
The WSTV model employs an anisotropic weighted matrix to the structure tensor total variation (STV) model, allowing it to better characterize local image features and maintain details during denoising.
The optimization problem of the WSTV model is solved using a fast first-order gradient projection algorithm, with a proven convergence rate of O(1/i^2).
Numerical experiments demonstrate that the WSTV model outperforms other TV-based methods, including TV, ATV, and STV, in terms of PSNR and SSIM for both grayscale and color image denoising, especially at high noise levels.
The WSTV model is more effective at restoring image details, such as edges and corners, compared to the STV model.
While the WSTV model performs well, it takes a relatively longer time compared to other methods. Future work could focus on improving the efficiency of the projection operators in the algorithm.
Stats
The WSTV model can effectively improve the quality of restored images compared to other TV and STV-based models.
The WSTV model exhibits a convergence rate of O(1/i^2).