Clearing Text Images: A Non-blind Deblurring with Convex Total Variation Regularization Model

In response to the escalating issue of obscured text in forensic analysis due to various factors such as illumination changes and low-resolution images, this paper introduces a novel convex penalty function within a total variation regularization model for text image deblurring. The proposed model enhances sparsity and edge preservation, yielding superior deblurring results. Accompanying this, an efficient Alternating direction method of multiplier algorithm is proposed, showcasing rapid convergence for complex deblurring tasks. Extensive experimentation validates the model's efficacy across diverse image resolutions, demonstrating superior performance in both qualitative and quantitative assessments, particularly in edge preservation and sparsity achievement. Additionally, the model exhibits superior numerical convergence compared to existing techniques, establishing its reliability for text image deblurring applications.