Abstract

Restricted by camera hardware, digital images captured by digital cameras are noisy, and the noise content of each color channel of the digital image is not balanced. However, most of the existing denoising algorithms assume that the entire image noise is constant, causing errors in the denoising of color images (non-uniform noise images), affecting image noise removal and texture detail protection. To solve this problem, we propose an evaluation operator that can describe the noise content and texture content in the local area of the image. According to the description value, the image pixels are classified, and heuristic denoising parameters are selected for each class to achieve a balance between noise removal effect and texture retention effect. Experimental results of multiple denoising methods show that the proposed algorithm has better denoising effect on color images.

Highlights

  • Most existing denoising methods are concentrated in additive white Gaussian noise (AWGN) [1]-[16], in which the observed noisy image is modeled as a composition of clean image and AWGN noise: z(i) x(i) n(i)

  • It is important to note that most of these methods assume that the noise variance of the entire image is fixed so that will inevitably bias the denoising result in the subsequent experiments, which will have a certain impact on the subsequent application

  • Gaussian noise among different color channels, and a Bayesian non-local mean denoising algorithm is employed in their paper to denoise images with non-uniform noise

Read more

Summary

Introduction

As a matter of fact, the noise level in the real noisy image is often non-uniform. Gaussian noise among different color channels, and a Bayesian non-local mean denoising algorithm is employed in their paper to denoise images with non-uniform noise. Xu et al [18] proposed a combined method which leverages a guided external prior and internal prior learning for non-uniform noise image denoising. In Chen et al [21], an adaptive BM3D filter was proposed to deal with non-uniform noise images, and in Plötz and Roth [22], a noise reduction algorithm for real photos was proposed. A new image denoising method within the framework of non-local means (NLM) regarding non-uniform noise is proposed.

Non-uniform Noise Model
Framework of NLM
Evaluation Operator for Noise Level and Texture Strength
Vote Strategy for Image Pixel Classification
Adaptive Setting of Denoising Parameters
Experimental Results and Analysis
Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call