Abstract:Noise removal in the standard RGB (sRGB) space remains a challenging task, in that the noise statistics of real-world images can be different in R, G and B channels. In fact, the green channel usually has twice the sampling rate in raw data and a higher signal-to-noise ratio than red/blue ones. However, the green channel prior (GCP) is often understated or ignored in color image denoising since many existing approaches mainly focus on modeling the relationship among image patches. In this paper, we propose a simple and effective one step GCP-based image denoising (GCP-ID) method, which aims to exploit the GCP for denoising in the sRGB space by integrating it into the classic nonlocal transform domain denoising framework. Briefly, we first take advantage of the green channel to guide the search of similar patches, which improves the patch search quality and encourages sparsity in the transform domain. Then we reformulate RGB patches into RGGB arrays to explicitly characterize the density of green samples. The block circulant representation is utilized to capture the cross-channel correlation and the channel redundancy. Experiments on both synthetic and real-world datasets demonstrate the competitive performance of the proposed GCP-ID method for the color image and video denoising tasks. The code is available at

What problem does this paper attempt to address?

The paper mainly addresses the problem of denoising color images, particularly the challenge of noise removal in the standard RGB (sRGB) color space. The authors propose a simple yet effective one-step denoising method based on the Green Channel Prior (GCP), named GCP-ID, which aims to leverage the characteristics of the green channel to enhance denoising performance. The paper points out that in real-world images, the noise statistical properties of different color channels may vary, with the green channel typically having a higher sampling rate and signal-to-noise ratio compared to the red or blue channels. However, most existing color image denoising methods primarily focus on modeling the relationships between image patches, neglecting the potential advantage of the green channel. Therefore, the proposed method integrates the green channel prior into the classical non-local transform domain denoising framework to fully exploit the advantages of the green channel for denoising. Specifically, the main contributions of the GCP-ID method include: 1. **Utilizing the green channel to guide similar image patch search**: Leveraging the low-noise characteristics of the green channel to improve the quality of similar image patch search, thereby encouraging sparsity in the transform domain. 2. **Reformulating image patches as RGGB arrays**: To explicitly represent the density and importance of green pixels, each RGB image patch is reformulated as an RGGB array with an additional green channel, and a block circulant matrix representation is used to capture inter-channel correlations. 3. **Non-local t-SVD transform**: Simplifying the model through a non-local t-SVD transform, effectively utilizing the redundancy between image patches and the correlations between channels. Experimental results show that the proposed GCP-ID method performs excellently on both synthetic and real-world datasets, achieving competitive performance for denoising tasks on color images and videos. Additionally, the method explores how the green channel prior can be extended to other imaging techniques, such as multispectral/hyperspectral imaging (MSI/HSI).

Color Image Denoising Using The Green Channel Prior

Image Denoising Using Green Channel Prior

QCGFNet: A Lightweight Color Image Denoising Network

Single image dehazing based on dark channel prior

Real-Color Image Denoised and Enhanced Synchronously Based on Wavelet Transform

A Brief Review of Real-World Color Image Denoising

Group Sparsity Residual Constraint for Image Denoising.

Signal-Dependent Noise Removal for Color Videos Using Temporal and Cross-Channel Priors

Group Sparsity Residual Constraint for Image Denoising with External Nonlocal Self-Similarity Prior

Multi-channel Weighted Nuclear Norm Minimization for Real Color Image Denoising

Image Denoising Using Hyper-Laplacian Priors and Gradient Histogram Preservation Model

A Generic Denoising Framework Via Guided Principal Component Analysis

Color image denoising based on multichannel non-local means fusion

Image denoising using group sparsity residual and external nonlocal self-similarity prior

Patch Group Based Nonlocal Self-Similarity Prior Learning for Image Denoising

External Patch Prior Guided Internal Clustering for Image Denoising

Cauchy Noise Removal Using Group-Based Low-Rank Prior

Multi-channel Nuclear Norm Minus Frobenius Norm Minimization for Color Image Denoising

A Comprehensive Comparison of Multi-Dimensional Image Denoising Methods

CID: Combined Image Denoising in Spatial and Frequency Domains Using Web Images

Rethink Gaussian Denoising Prior for Real-World Image Denoising