Unified Diffusion Smoothing Models for Edge-Preserving Image Denoising Amidst Gaussian and Mixed Noise

In image denoising, particularly under the influence of Gaussian and mixed noise, the challenge of preserving edge integrity while eliminating noise is paramount. This is largely due to the tendency of Gaussian noise removal techniques to induce edge blurring. Within this context, diffusion smoothing algorithms emerge as a potent solution, offering the dual benefits of image smoothing and edge preservation. The present study conducts a comprehensive review of four foundational diffusion smoothing algorithms and introduces a novel, unified model for the diffusion algorithm class. This model posits that any diffusion function fundamentally relies on a statistical estimation operation, such as mean, weighted mean, median, mode, and adaptive weighted mean, among others. Consequently, existing diffusion models can be reinterpreted through this unified framework, facilitating the development of new models aimed at enhancing filter performance and reducing computational complexity. Adhering to the unified model, four innovative diffusion smoothing models were formulated. The performance of these models was subjected to both qualitative analysis and evaluation based on standard performance metrics. Results demonstrate that the proposed models maintain satisfactory performance levels, even in scenarios characterized by high noise intensities, outperforming traditional diffusion models. This study underscores the versatility and efficacy of the unified model in refining image denoising techniques, thereby contributing significantly to the field of image processing.