BlindPSNR is proposed, a lightweight no-reference network that fuses the enhanced image with the degraded low-light input via windowed cross-attention and estimates PSNR through heteroscedastic regression, providing exact supervision without a separate teacher network.
Abstract
Low-light image enhancement (LLIE) methods involve tunable parameters that are typically fixed, often leading to performance degradation when applied across scenes. Manually selecting the best configuration, however, can be time-consuming and not always practical. Peak signal-to-noise ratio (PSNR) is the natural fidelity criterion for automating parameter selection, yet it requires a ground-truth reference that is typically unavailable. To our knowledge, no learning-based method addresses no-reference PSNR prediction for low-light image enhancement; the natural surrogate, no-reference image quality assessment (NR-IQA), targets perceptual quality rather than signal fidelity, and all seven baselines we test achieve 0% top-1 selection accuracy on our benchmark. With paired training data, the ground-truth PSNR is analytically computable, providing exact supervision without a separate teacher network. Building on this, we propose BlindPSNR, a lightweight no-reference network that fuses the enhanced image with the degraded low-light input via windowed cross-attention and estimates PSNR through heteroscedastic regression. While a scalar-regression baseline achieves top-1 accuracy of 54.4%, BlindPSNR raises this to 89.5% with regret dropping from 1.62 dB to 0.026 dB, and generalizes to unseen datasets (SRCC = 0.61-0.67).
The results position LCR-Net as a lightweight controlled-enhancement alternative that emphasizes residual-darkness reduction, restrained exposure behavior, and computational efficiency rather than unrestricted reconstruction accuracy.
Tao Jiang, Su-Hang Yang, Yu-Chen Lu et al.· Journal of King Saud Univers...· 0 citations
Low-light image enhancement (LLIE) requires more than brightness amplification because real low-light images often exhibit spatially varying illumination, amplified noise, color distortion, and local detail loss. Recent SSM- and Mamba-based restorers provide efficient long-range modeling, yet the route for introducing...
Wei-Chen Ng, Qi-Yi Xie, Jian-Ding Tan et al.· 0 citations
A Dynamic Interpolation Training-only Module (DITM), which introduces no additional learnable parameters and adaptively reshapes the model prediction for supervised feed-forward LLIE training, and improves fidelity and perceptual quality while preserving the inference-time architecture and introducing no additional inf...
Pei-Rui Zhu, Shi-Jie Hao, Jing-Xi Liao et al.· IEEE transactions on circuit...· 0 citations
The proposed FreqLight-Net is a five-million-parameter U-Net augmented with a Fast Fourier Transform branch at the bottleneck, where spatial and spectral features are fused through a learnable convex combination before decoding, suggesting that frequency-aware compact U-Nets offer a deployable path for real-time low-li...
M. Al-Azzawi, H. I. Wahhab, Safaa Aqeel Khazraji et al.· JOIV: International Journal...· 0 citations
This paper couples an iterative curve estimator with a refinement module conditioned on the enhancement process itself, which attains a higher PSNR than the rational mapping the authors adopt, so it is presented as a characterised design choice, not a demonstrated optimum.
A novel Hierarchical Multi-Scale Cross-Attention Network that effectively captures both local distortion patterns and global semantic information for quality prediction and exhibits superior generalization capability compared to existing approaches is proposed.
Jiakuo Yan, Jun Zeng· International Conference on...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.