Multi-objective Optimization of Anisotropic Diffusion Parameters for Enhanced Image Denoising

Noise often corrupts images during their acquisition, storage, and transmission processes, resulting in a degradation of image quality. The main challenge in image denoising is to remove the corrupted information while preserving the fine details of the original image. Anisotropic Diffusion (AD) is a well-established method for noise removal in digital images that effectively preserves the image edges. However, selecting the appropriate parameters for the AD operation significantly influences the filtering outcomes. Despite its importance, the automatic determination of AD parameters based on image requirements has received limited attention in the literature. This chapter introduces a novel multi-objective approach for obtaining optimal AD parameters that yield effective filtering results. The presented methodology aims to strike the best possible balance between two conflicting objectives: minimizing noise content in the image while maximizing contrast. Improving one objective without compromising the other is not feasible. To address this, the presented approach employs the Non-dominated Sorting Genetic Algorithm based on reference points (NSGA-III), which is known as a robust and powerful algorithm for multi-objective optimization. The final solution is determined by analyzing the Optimal Pareto front. Experimental results demonstrate that the presented method outperforms existing filtering algorithms in terms of both visual quality and standard performance metrics.