HIGH-FIDELITY SOLAR SUNSPOT IMAGE DENOISING: A COMPARATIVE ANALYSIS OF ADAPTIVE ANISOTROPIC DIFFUSION, NON-LOCAL MEANS, AND HYBRID FILTERING APPROACHES

Abstract

Solar image denoising plays a critical role in enhancing sunspot detection accuracy, solar activity monitoring, and the predictive capabilities of space weather models. Nevertheless, data captured by solar observatories are often corrupted by various forms of degradation, including sensor, photon, and speckle noise, as well as instrumental artifacts, which collectively diminish image fidelity and hinder automated analysis pipelines. To mitigate these issues, this research evaluates three distinct denoising strategies: Adaptive Anisotropic Diffusion, Non-Local Means, and a hybrid AAD+NLM approach. By integrating AAD’s edge-preserving diffusion capabilities with NLM’s non-local similarity analysis, the proposed hybrid framework achieves superior noise suppression while maintaining critical structural features. The effectiveness of these methodologies was assessed utilizing SDO/HMI, AIA 171 Å, and AIA 193 Å datasets, with performance evaluated through metrics such as Peak Signal-to-Noise Ratio, Structural Similarity Index, Mean Squared Error, Edge Preservation Index, and computational efficiency. Experimental findings indicate that the hybrid approach yields superior denoising performance and enhanced structural integrity, resulting in more robust sunspot visualization and confirming its efficacy for automated solar image analysis.