Image Denoising Based on Deep Image Prior Combined Sparsity with Regularization by Denoising

Image restoration is a critical technique for enhancing image quality, widely applied in fields such as meteorological remote sensing, medical diagnosis, and digital photography. This article delves into the challenges of image restoration, highlighting the ill-posed nature of the problem and the limitations of traditional regularization methods. It introduces a novel image denoising model that combines deep image prior (DIP), regularization by denoising (RED), and the norm as explicit regularizers. This innovative approach leverages the generative capabilities of DIP to capture fine image details, while RED and the norm enhance denoising performance and robustness. The article presents a detailed description of the new model and its algorithm, including the use of the alternating direction method of multipliers (ADMM) for efficient optimization. Experimental results demonstrate the superior performance of the proposed method compared to existing state-of-the-art techniques, showcasing improved peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) metrics. The article also discusses the flexibility and adaptability of the new model, making it a promising solution for various image restoration tasks, particularly in scenarios where training datasets are limited.