Image Denoising using Adaptive NL Means Filtering with Method Noise Thresholding

Abstract

Background/Objectives: Image denoising is an important step in image processing applications. Usually noise is added to the original image during transmission, acquisition and storage process and is considered as noisy image. For precise analysis and extraction of image features, the noisy image is denoised without losing the original image details. This study aims to introduce a novel denoising method to obtain denoised image(s) such that it has fewer artifacts and is more efficient at higher noise levels. Method: The proposed novel denoising method introduces Adaptive Non Local Means (ANL) along with Method Noise Thresholding (MNT) technique to improve the image quality of the denoised image. Method Noise (MN) image obtained by taking the difference of image details between noisy image and pre-filtered mage. Recovered value from the MN through thresholding includes some of the important components of the original image. These values computed added to pre-filtered image to recover image features of the original image. Findings: The standard image, denoised with noise standard (σ =10) using bior6.8 wavelet when filtered using existing Gaussian Bilateral Filter along with Method-Noise Thresholding filtering technique and Wiener Filter along with Residual Thresholding show improvement in quality of the denoised image in terms of PSNR and ISSN values as compared to the proposed filtering technique. The proposed filter technique results in higher PSNR and ISSN values (PSNR =33.80 and SSIN =0.9994). Novelty: It is known that ANLM results in improved denoised parameters compared with NLM filter; however, when MNT is blended with ANLM shows further improvement in quality of denoised image. Hence, in the proposed method, MNT is incorporated along with ANLM for improvement in denoising process. Image Quality Index (IQI) of the different standard images using ANLMT filtering technique is also studied.

Image Denoising Techniques Using Unsupervised Machine Learning and Deep Learning Algorithms: A Review

“…Additionally, statistical modeling techniques, such as Bayesian methods [30] or non-local means [31], have gained prominence for their ability to exploit statistical priors and image self-similarity to enhance denoising performance. These methods leverage patch-based approaches, comparing similar patches within an image to estimate and reduce noise while preserving fine details.…”

“…In [31] the unique denoising method uses Adaptive Non-Local Means (ANL) and Method Noise Thresholding (MNT) to improve image quality. Method Noise (MN) image is the difference between noisy and pre-filtered image information.…”