Official implementation of "A unified Benchmark for Multi-Frame Image Restoration under Severe Refractive Warping" (CVPR 2026) This project introduces a benchmark for evaluating multi-frame image reconstruction methods under severe refrctive warping conditions at air-water interface, a challenging distortion regime characterized by spatially varying geometric deformation across frames. The benchmark enables controlled and reproducible evaluation by generating distorted image sequences procedurally from wave profiles for a specific wave type and background images provided with abovelisted dataset.
| argument | description |
|---|---|
| --dataset_root | root directory of the above dataset with backgrounds and wave profiles. |
| --wave_type | wave type ocean/shallow/sine/ripple |
| --amplitude | wave amplitude low/mid/high/extreme |
| --L | number of frames in eval |
| --method | evaluated method ref/mean/grid_registration/datum |
python eval.py --dataset_root ${your_data_path} --L 1 --wtype ocean --amplitude low --method ref
python eval.py --dataset_root ${your_data_path} --L 49 --wtype ripple --amplitude mid --method datum
Download the provided dataset for wave profiles and backgrounds
Please consider citing our work as follows if it is helpful.