Michele Claus, Jan van Gemert
We propose ViDeNN: a CNN for Video Denoising without prior knowledge on the noise distribution (blind denoising). The CNN architecture uses a combination of spatial and temporal filtering, learning to spatially denoise the frames first and at the same time how to combine their temporal information, handling objects motion, brightness changes, low-light conditions and temporal inconsistencies. We demonstrate the importance of the data used for CNNs training, creating for this purpose a specific dataset for low-light conditions. We test ViDeNN on common benchmarks and on self-collected data, achieving good results comparable with the state-of-the-art.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Denoising | CBSD68 sigma5 | PSNR | 39.73 | Spatial-CNN |
| Denoising | CBSD68 sigma15 | PSNR | 33.66 | Spatial-CNN |
| Denoising | CBSD68 sigma25 | PSNR | 30.99 | Spatial-CNN |
| Denoising | CBSD68 sigma35 | PSNR | 29.34 | Spatial-CNN |
| Denoising | CBSD68 sigma10 | PSNR | 35.92 | Spatial-CNN |
| Denoising | CBSD68 sigma50 | PSNR | 27.63 | Spatial-CNN |
| 3D Architecture | CBSD68 sigma5 | PSNR | 39.73 | Spatial-CNN |
| 3D Architecture | CBSD68 sigma15 | PSNR | 33.66 | Spatial-CNN |
| 3D Architecture | CBSD68 sigma25 | PSNR | 30.99 | Spatial-CNN |
| 3D Architecture | CBSD68 sigma35 | PSNR | 29.34 | Spatial-CNN |
| 3D Architecture | CBSD68 sigma10 | PSNR | 35.92 | Spatial-CNN |
| 3D Architecture | CBSD68 sigma50 | PSNR | 27.63 | Spatial-CNN |