Image Denoising using Traditional and Attention Based Techniques: A Survey

In this paper, the goal is to explore models for reducing noise in images while preserving important details like textures and edges. The paper covers various methods used for image denoising, ranging from traditional approaches like Block Matching and 3D Filtering (BM3D) and Principal Component Analysis (PCA) to more advanced methods using deep learning. The focus is on how newer models like Autoencoders, Convolutional Neural Networks (CNNs), and have improved denoising by learning patterns in noisy images. The paper also highlights cuttingedge models like Vision Transformers (ViTs) and Denoising Convolutional Neural Networks (DnCNNs), which use innovative ways to analyze and clean images by treating images as sequences or patches, helping achieve better results. Advancement in deep learning has enhanced the ability of the denoising the images while maintaining crucial details. Models like ViTs and DnCNNs have become the most widely used solutions in order to balance noise reduction.