International Journal of Science and Technology (IJST)

A Review of Image Denoising Techniques: Models, Learning Paradigms and Emerging Trends

Image denoising is a cornerstone problem in digital image processing and computer vision, aiming to remove noise while preserving essential image structures such as edges and textures. Over the years, denoising methods have evolved from classical spatial and transform-domain filters to sophisticated learning-based and deep neural network models. This paper presents a comprehensive and critical review of image denoising techniques, emphasizing methodological evolution, underlying assumptions and practical limitations. Unlike conventional surveys that focus primarily on algorithmic categorization, this review highlights geometric adaptivity, noise modelling realism and learning paradigms as key dimensions shaping modern denoising research. We analysed traditional, model-based and deep learning approaches, discussed their strengths, weaknesses and identified open challenges such as real-world noise generalization, edge preservation and interpretability. Finally, emerging trends and future research directions are outlined, positioning image denoising as a continually evolving field driven by both theoretical advances and real-world imaging demands.