Published Paper Details:

The paper introduces a novel framework called Nonlocal Means based Framework (NMF) for image denoising. The goal is to reduce noise in digital images without compromising important image features. While existing methods often focus on removing either additive white Gaussian noise (AWGN) or impulse noise (IN) separately, this framework addresses the challenge of images corrupted by a mixture of both types. The proposed framework combines the strengths of the median filter and nonlocal means to tackle the noise mixture. The median filter, a non-linear filtering technique, is employed to identify outlier pixels in the image, which are then replaced by nonlocal means. This step effectively separates the mixed noise into its Gaussian components. To further enhance denoising performance, the framework utilizes a low-rank approximation combined with NMF (LRNM) model. This model groups similar nonlocal patches into a matrix and applies a low-rank approximation to reconstruct the clean image. Additionally, a Convolutional Neural Network (CNN) is integrated with NMF (NMF-CNN) to demonstrate the versatility of the NMF approach. The experimental results highlight the effectiveness of LRNM and NMF-CNN in removing mixed noise and producing visually pleasing denoised images. These approaches demonstrate strong performance in noise reduction and can contribute to improving the quality of images captured by modern cameras.