Abstract:Recently, multi-modal image processing has shown its great potential in boosting denoising performance in terms of both accuracy and visual quality. However, current studies generally face the challenge of achieving a good balance between noise removal and detail preservation. In this work, we introduce patch-wise frequency decomposition into convolutional neural networks and propose a novel multi-modal image denoising (MID) algorithm. Integrating the frequency-domain information of images from different modalities, our network first predicts a frequency-relevant residual and then regresses the denoised result using a learnable reconstruction kernel. Benefiting from the distinctive properties of noise and true signals as well as the correlation between multi-modal images in the frequency domain, the proposed algorithm can effectively remove noise and simultaneously reconstruct fine details. Extensive experiments demonstrate the superiority and generalizability of our algorithm over state-of-the-art competing algorithms on various MID tasks, including near-infrared guided RGB image denoising, flash guided no-flash image denoising, and RGB guided depth image denoising. Code is available at https://github.com/liuxw11/FRL.

Local And Non-Local Means Based Mixed Filtering For Video Sequences Denoising

Adaptive fuzzy filter algorithm for real-time video denoising

An Improved Non-local Means Image De-noising Algorithm Using Mahalanobis Distance

Frequency-Relevant Residual Learning for Multi-Modal Image Denoising.

DENOISING METHOD BASED ON WAVELET TRANSFORM COEFFICIENT CORRELATION AND LOCAL LAPLACIAN MODEL

A Non-Local Means Filter for Removing the Poisson Noise

Fast Non-Local Algorithm for Image Denoising

A New Method for Nonlocal Means Image Denoising Using Multiple Images

Exploiting Image Local and Nonlocal Consistency for Mixed Gaussian-Impulse Noise Removal

Adaptive Image Denoising Method Based on Non-Local Means Filtering

A Robust and Fast Non-Local Means Algorithm for Image Denoising

Multichannel Nonlocal Means Fusion for Color Image Denoising

A Novel Segmentation Based Video-Denoising Method with Noise Level Estimation

Illumination-invariant non-local means based video denoising

Linear Attention Based Deep Nonlocal Means Filtering for Multiplicative Noise Removal

A New Image Filtering Method: Nonlocal Image Guided Averaging

Mixed Gaussian-impulse Video Noise Removal Via Temporal-Spatial Decomposition.

Image Denoising Based On A Cnn Model

Non-local Means Image Denoising in Wavelet Domain with Unknown Noise Variance

3-D nonlocal means filter with noise estimation for hyperspectral imagery denoising