595 pairs of real-world low-resolution and high-resolution images captured using Canon 5D Mark IV and Nikon D810 DSLR cameras. The dataset provides ground truth for super-resolution tasks across x2, x3, and x4 scaling factors, utilizing optical zoom rather than synthetic degradation.
Use Cases
- Train single-image super-resolution models using the LR and HR image pairs to handle real-world sensor noise.
- Benchmark the performance of image restoration algorithms against the x2, x3, and x4 scaling subsets.
- Develop and test image registration techniques using the raw, unaligned captures from the DSLR source files.
Strengths
- Contains 595 image pairs across two different camera brands: Canon and Nikon.
- Includes ground truth data for three specific scaling factors: x2, x3, and x4.
- Features pixel-aligned image pairs processed through a dedicated image registration algorithm.
- Covers diverse indoor and outdoor scenes to provide varied noise and blur profiles.