Gradient Variance Loss for Structure-Enhanced Image Super-Resolution

02/02/2022
by   Lusine Abrahamyan, et al.
0

Recent success in the field of single image super-resolution (SISR) is achieved by optimizing deep convolutional neural networks (CNNs) in the image space with the L1 or L2 loss. However, when trained with these loss functions, models usually fail to recover sharp edges present in the high-resolution (HR) images for the reason that the model tends to give a statistical average of potential HR solutions. During our research, we observe that gradient maps of images generated by the models trained with the L1 or L2 loss have significantly lower variance than the gradient maps of the original high-resolution images. In this work, we propose to alleviate the above issue by introducing a structure-enhancing loss function, coined Gradient Variance (GV) loss, and generate textures with perceptual-pleasant details. Specifically, during the training of the model, we extract patches from the gradient maps of the target and generated output, calculate the variance of each patch and form variance maps for these two images. Further, we minimize the distance between the computed variance maps to enforce the model to produce high variance gradient maps that will lead to the generation of high-resolution images with sharper edges. Experimental results show that the GV loss can significantly improve both Structure Similarity (SSIM) and peak signal-to-noise ratio (PSNR) performance of existing image super-resolution (SR) deep learning models.

READ FULL TEXT

page 1

page 4

research
03/29/2020

Structure-Preserving Super Resolution with Gradient Guidance

Structures matter in single image super resolution (SISR). Recent studie...
research
12/18/2018

SREdgeNet: Edge Enhanced Single Image Super Resolution using Dense Edge Detection Network and Feature Merge Network

Deep learning based single image super-resolution (SR) methods have been...
research
01/18/2019

Learning a Deep Convolution Network with Turing Test Adversaries for Microscopy Image Super Resolution

Adversarially trained deep neural networks have significantly improved p...
research
06/19/2023

Evaluating Loss Functions and Learning Data Pre-Processing for Climate Downscaling Deep Learning Models

Deep learning models have gained popularity in climate science, followin...
research
07/23/2020

Frequency Domain-based Perceptual Loss for Super Resolution

We introduce Frequency Domain Perceptual Loss (FDPL), a loss function fo...
research
06/26/2023

Iterative-in-Iterative Super-Resolution Biomedical Imaging Using One Real Image

Deep learning-based super-resolution models have the potential to revolu...
research
02/15/2023

Enhancing Biogenic Emission Maps Using Deep Learning

Biogenic Volatile Organic Compounds (BVOCs) play a critical role in bios...

Please sign up or login with your details

Forgot password? Click here to reset