IDENet-LF: Implicit Degradation Estimation for Blind Light-Field Super-Resolution

Abstract

We propose IDENet-LF, a blind light-field spatial super-resolution (SR) framework that replaces the explicit degradation estimator used in prior LF blind SR pipelines with the lightweight implicit degradation estimation network (IDENet). By integrating IDENet as the first module and retaining the original light-field restoration (LF-SR) backbone from LF-DEST, the proposed system jointly benefits from IDENet’s efficient implicit degradation representation and LF-DEST’s spatial-angular restoration capabilities. The framework is tested on public light-field benchmarks under diverse degradations, shows improved robustness to unknown blur and noise kernels. Extensive experiments on multiple LF benchmarks demonstrate that IDENet-LF achieves comparable performance to LF-DEST, improving PSNR by up to 0.2 dB. The results confirm that implicit degradation modeling is a promising direction for efficient and robust blind LF super-resolution.