However, we observe that their framework for image degradation is not ideal for SR of LR face images from surveillance cameras, as these are often also corrupted by compression artifacts. who propose to perform RWSR via kernel estimation and noise injection. This approach is inspired by the work of Ji et al. Creating LR images from clean high-quality images, that is, the target domain, allows us to train an SR model that learns to super-resolve images with similar characteristics. To create an SR model that is robust against the corruptions found in real images, we create a degradation framework that can produce LR images that have the same image characteristic as the images that we want to super-resolve, that is, the source domain images. We apply our proposed method to face images, but the method is also applicable to other image domains. This paper is about SR of real low-resolution, noisy, and corrupted images, also known as Real-World Super-Resolution (RWSR). Hence, when using algorithms trained with supervised learning on such artificial LR/HR image pairs, the reconstructed images usually contains strong artifacts due to the domain gap. This is mostly due to the fact that the downsampling operation with bicubic downscaling changes the natural image characteristics and reduces the amount of artifacts.

However, these methods often assume that the LR images were downsampled with bicubic interpolation, and therefore fail to produce good results when applied to real-world LR images. Current SoTA face SR methods based on Convolutional Neural Networks (CNNs) are able to reconstruct images with photo-realistic appearance from artificially generated LR images. This is useful in many different applications such as video surveillance and face enhancement. In addition, it is shown that the traditional non-reference Image Quality Assessment (IQA) methods fail to capture this improvement and demonstrate that the more recent NIMA metric correlates better with human perception via Mean Opinion Rank (MOR).įace SR is a special case of SR which aims to restore HR face images from their LR counterparts.

Experimental results on both real and artificially corrupted face images show that our method results in more detailed reconstructions with less noise compared to the existing State-of-the-Art (SoTA) methods. For better perceptual quality, a Generative Adversarial Network (GAN) based SR model is used, where the commonly used VGG-loss is exchanged with LPIPS-loss.

This allows to train an SR model using high-quality face images as Ground-Truth (GT). The framework estimates realistic blur kernels, noise distributions, and JPEG compression artifacts to generate LR images with similar image characteristics as the ones in the source domain. To solve this problem, a novel framework for the generation of realistic LR/HR training pairs is proposed. Hence, SR methods trained on such data most often fail to produce good results when applied to real LR images. This operation changes the natural image characteristics and reduces noise. Most existing face image Super-Resolution (SR) methods assume that the Low-Resolution (LR) images were artificially downsampled from High-Resolution (HR) images with bicubic interpolation.

IET Generation, Transmission & Distribution.

IET Electrical Systems in Transportation.

IET Cyber-Physical Systems: Theory & Applications.

IET Collaborative Intelligent Manufacturing.

CAAI Transactions on Intelligence Technology.

Arma 3 face generator