RobustSCI: Beyond Reconstruction to Restoration for Snapshot Compressive Imaging under Real-World Degradations

TL;DR

This work pioneers the first study on robust video SCI restoration, shifting the goal from construction to restoration--recovering the underlying pristine scene from a degraded measurement, and introduces RobustSCI-C (RobustSCI-Cascade), which integrates a pre-trained Lightweight Post-processing Deblurring Network to significantly boost restoration performance with minimal overhead.

Abstract

Deep learning algorithms for video Snapshot Compressive Imaging (SCI) have achieved great success, yet they predominantly focus on reconstructing from clean measurements. This overlooks a critical real-world challenge: the captured signal itself is often severely degraded by motion blur and low light. Consequently, existing models falter in practical applications. To break this limitation, we pioneer the first study on robust video SCI restoration, shifting the goal from "reconstruction" to "restoration"--recovering the underlying pristine scene from a degraded measurement. To facilitate this new task, we first construct a large-scale benchmark by simulating realistic, continuous degradations on the DAVIS 2017 dataset. Second, we propose RobustSCI, a network that enhances a strong encoder-decoder backbone with a novel RobustCFormer block. This block introduces two parallel branches--a multi-scale deblur branch and a frequency enhancement branch--to explicitly disentangle and remove degradations during the recovery process. Furthermore, we introduce RobustSCI-C (RobustSCI-Cascade), which integrates a pre-trained Lightweight Post-processing Deblurring Network to significantly boost restoration performance with minimal overhead. Extensive experiments demonstrate that our methods outperform all SOTA models on the new degraded testbeds, with additional validation on real-world degraded SCI data confirming their practical effectiveness, elevating SCI from merely reconstructing what is captured to restoring what truly happened.

Figures (6)

• Figure 1: PSNR performance of our methods against SOTA models across 10 scenarios (clean, 3 motion blur (MB), 3 low-light (LL), 3 mixed degradation levels) on grayscale benchmarks. Our models demonstrate superior robustness as degradation severity increases.
• Figure 2: Overview of our training and inference pipeline. Paired data is generated by applying simulated degradations to clean video frames, followed by CACTI simulation to produce degraded measurements (${{\boldsymbol Y}}$). RobustSCI is trained end-to-end to recover clean ground truth frames (${\boldsymbol X}$), while the Lightweight Post-processing Deblurring Network (trained separately) can be used as a frozen post-processing step during inference.
• Figure 3: Architecture of the proposed RobustCFormer block. Input features are processed via four parallel branches (ST-Baseline: SCB+TSAB, MSDB, FEB) before fusion. Aggregated features are refined by an FFN with a residual connection.
• Figure 4: Qualitative comparison on the grayscale runner benchmark under different degradation scenarios. From top to bottom, the rows show results for Clean, MotionBlur-L2 (MB-L2), LowLight-L2 (LL-L2), and Mixed-L2 scenarios. The first column displays the ground truth or the degraded input. Each subsequent column shows the reconstruction from a different model.
• Figure 5: Qualitative comparison on the color Jockey benchmark. From top to bottom, the rows show the input and the results from different models. From left to right, the columns show different degradation scenarios at the L2 level.