# Super Resolution using Convolutional Neural Networks

This module contains several learning-based algorithms for upscaling an image.

## Usage

Run the following command to build this module:

```make
cmake -DOPENCV_EXTRA_MODULES_PATH=<opencv_contrib>/modules -Dopencv_dnn_superres=ON <opencv_source_dir>
```

Refer to the tutorials to understand how to use this module.

## Models

There are four models which are trained.

#### EDSR

Trained models can be downloaded from [here](https://github.com/Saafke/EDSR_Tensorflow/tree/master/models).

- Size of the model: ~38.5MB. This is a quantized version, so that it can be uploaded to GitHub. (Original was 150MB.)
- This model was trained for 3 days with a batch size of 16
- Link to implementation code: https://github.com/Saafke/EDSR_Tensorflow
- x2, x3, x4 trained models available
- Advantage: Highly accurate
- Disadvantage: Slow and large filesize
- Speed: < 3 sec for every scaling factor on 256x256 images on an Intel i7-9700K CPU.
- Original paper: [Enhanced Deep Residual Networks for Single Image Super-Resolution](https://arxiv.org/pdf/1707.02921.pdf) [1]

#### ESPCN

Trained models can be downloaded from [here](https://github.com/fannymonori/TF-ESPCN/tree/master/export).

- Size of the model: ~100kb
- This model was trained for ~100 iterations with a batch size of 32
- Link to implementation code: https://github.com/fannymonori/TF-ESPCN
- x2, x3, x4 trained models available
- Advantage: It is tiny and fast, and still performs well.
- Disadvantage: Perform worse visually than newer, more robust models.
- Speed: < 0.01 sec for every scaling factor on 256x256 images on an Intel i7-9700K CPU.
- Original paper: [Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network](https://arxiv.org/pdf/1707.02921.pdf) [2]

#### FSRCNN

Trained models can be downloaded from [here](https://github.com/Saafke/FSRCNN_Tensorflow/tree/master/models).

- Size of the model: ~40KB (~9kb for FSRCNN-small)
- This model was trained for ~30 iterations with a batch size of 1
- Link to implementation code: https://github.com/Saafke/FSRCNN_Tensorflow
- Advantage: Fast, small and accurate
- Disadvantage: Not state-of-the-art accuracy
- Speed: < 0.01 sec for every scaling factor on 256x256 images on an Intel i7-9700K CPU.
- Notes: FSRCNN-small has fewer parameters, thus less accurate but faster.
- Original paper: [Accelerating the Super-Resolution Convolutional Neural Network](http://mmlab.ie.cuhk.edu.hk/projects/FSRCNN.html) [3]

#### LapSRN

Trained models can be downloaded from [here](https://github.com/fannymonori/TF-LapSRN/tree/master/export).

- Size of the model: between 1-5Mb
- This model was trained for ~50 iterations with a batch size of 32
- Link to implementation code: https://github.com/fannymonori/TF-LAPSRN
- x2, x4, x8 trained models available
- Advantage: The model can do multi-scale super-resolution with one forward pass. It can now support 2x, 4x, 8x, and [2x, 4x] and [2x, 4x, 8x] super-resolution.
- Disadvantage: It is slower than ESPCN and FSRCNN, and the accuracy is worse than EDSR.
- Speed: < 0.1 sec for every scaling factor on 256x256 images on an Intel i7-9700K CPU.
- Original paper: [Deep laplacian pyramid networks for fast and accurate super-resolution](https://arxiv.org/pdf/1707.02921.pdf) [4]

### Benchmarks

Comparing different algorithms. Scale x4 on monarch.png (768x512 image).

|               | Inference time in seconds (CPU)| PSNR | SSIM |
| ------------- |:-------------------:| ---------:|--------:|
| ESPCN            |0.01159   | 26.5471 | 0.88116 |
| EDSR             |3.26758     |**29.2404**  |**0.92112**  |
| FSRCNN           | 0.01298   | 26.5646 | 0.88064 |
| LapSRN           |0.28257    |26.7330   |0.88622  |
| Bicubic          |0.00031 |26.0635  |0.87537  |
| Nearest neighbor |**0.00014** |23.5628  |0.81741  |
| Lanczos          |0.00101  |25.9115  |0.87057  |

Refer to the benchmarks located in the tutorials for more detailed benchmarking.

### References
[1] Bee Lim, Sanghyun Son, Heewon Kim, Seungjun Nah, and Kyoung Mu Lee, **"Enhanced Deep Residual Networks for Single Image Super-Resolution"**, <i> 2nd NTIRE: New Trends in Image Restoration and Enhancement workshop and challenge on image super-resolution in conjunction with **CVPR 2017**. </i> [[PDF](http://openaccess.thecvf.com/content_cvpr_2017_workshops/w12/papers/Lim_Enhanced_Deep_Residual_CVPR_2017_paper.pdf)] [[arXiv](https://arxiv.org/abs/1707.02921)] [[Slide](https://cv.snu.ac.kr/research/EDSR/Presentation_v3(release).pptx)]

[2] Shi, W., Caballero, J., Huszár, F., Totz, J., Aitken, A., Bishop, R., Rueckert, D. and Wang, Z., **"Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network"**, <i>Proceedings of the IEEE conference on computer vision and pattern recognition</i> **CVPR 2016**. [[PDF](http://openaccess.thecvf.com/content_cvpr_2016/papers/Shi_Real-Time_Single_Image_CVPR_2016_paper.pdf)] [[arXiv](https://arxiv.org/abs/1609.05158)]

[3] Chao Dong, Chen Change Loy, Xiaoou Tang. **"Accelerating the Super-Resolution Convolutional Neural Network"**, <i> in Proceedings of European Conference on Computer Vision </i>**ECCV 2016**. [[PDF](http://personal.ie.cuhk.edu.hk/~ccloy/files/eccv_2016_accelerating.pdf)]
[[arXiv](https://arxiv.org/abs/1608.00367)] [[Project Page](http://mmlab.ie.cuhk.edu.hk/projects/FSRCNN.html)]

[4] Lai, W. S., Huang, J. B., Ahuja, N., and Yang, M. H., **"Deep laplacian pyramid networks for fast and accurate super-resolution"**, <i> In Proceedings of the IEEE conference on computer vision and pattern recognition </i>**CVPR 2017**. [[PDF](http://openaccess.thecvf.com/content_cvpr_2017/papers/Lai_Deep_Laplacian_Pyramid_CVPR_2017_paper.pdf)] [[arXiv](https://arxiv.org/abs/1710.01992)] [[Project Page](http://vllab.ucmerced.edu/wlai24/LapSRN/)]