Abstract:Human face captured at night or in dimly lit environments has become a common practice, accompanied by complex low-light and low-resolution degradations. However, the existing face super-resolution (FSR) technologies and derived cascaded schemes are inadequate to recover credible textures. In this paper, we propose a novel approach that decomposes the restoration task into face structural fidelity maintaining and texture consistency learning. The former aims to enhance the quality of face images while improving the structural fidelity, while the latter focuses on eliminating perturbations and artifacts caused by low-light degradation and reconstruction. Based on this, we develop a novel low-light low-resolution face super-resolution framework. Our method consists of two steps: an illumination correction face super-resolution network (IC-FSRNet) for lighting the face and recovering the structural information, and a detail enhancement model (DENet) for improving facial details, thus making them more visually appealing and easier to analyze. As the relighted regions could provide complementary information to boost face super-resolution and vice versa, we introduce the mutual learning to harness the informative components from relighted regions and reconstruction, and achieve the iterative refinement. In addition, DENet equipped with diffusion probabilistic model is built to further improve face image visual quality. Experiments demonstrate that the proposed joint optimization framework achieves significant improvements in reconstruction quality and perceptual quality over existing two-stage sequential solutions. Code is available at https://github.com/wcy-cs/IC-FSRDENet.

Low-Light Face Super-resolution Via Illumination, Structure, and Texture Associated Representation

Deep Illumination-Enhanced Face Super-Resolution Network for Low-Light Images

Research on Low-Light Image Enhancement Algorithm Based on Filtering and Data-Driven Composite

Face Super-Resolution Via Progressive-Scale Boosting Network

Learning deep texture-structure decomposition for low-light image restoration and enhancement

Facial mask attention network for identity-aware face super-resolution

The face image super-resolution algorithm based on combined representation learning

Noisy Practical Facial Super-Resolution Method Via Deformable Constrained Model with Small Dataset

Watch You under Low-resolution and Low-illumination: Face Enhancement Via Bi-factor Degradation Decoupling

A Face Structure Attention Network for Face Super-Resolution

Multi-Scale Feature Fusion and Structure-Preserving Network for Face Super-Resolution

Robust Face Image Super-Resolution via Joint Learning of Subdivided Contextual Model

Efficient Face Super-Resolution via Wavelet-based Feature Enhancement Network

Iterative Attribute Augmentation Network for Face Image Super Resolution

Facial Attribute Capsules for Noise Face Super Resolution

Revisiting coarse-to-fine strategy for low-light image enhancement with deep decomposition guided training

Deep Face Super-Resolution with Iterative Collaboration between Attentive Recovery and Landmark Estimation

A joint learning method with consistency-aware for low-resolution facial expression recognition

FRC-Net: A Simple Yet Effective Architecture for Low-Light Image Enhancement

Face Super-Resolution with Progressive Embedding of Multi-scale Face Priors

A Unified Framework to Super-Resolve Face Images of Varied Low Resolutions