Multi-Scale Recursive and Perception-Distortion Controllable Image Super-Resolution
This work addresses the challenge of balancing perceptual quality and distortion in image super-resolution for applications like photography and computer vision, though it is incremental as it builds on existing methods.
The paper tackled the problem of image super-resolution by developing a modified Multi-Grid Back-Projection architecture with a controllable parameter for artificial details, achieving 2nd best perceptual quality for average RMSE<=16, 5th for RMSE<=12.5, and 7th for RMSE<=11.5 in the PIRM 2018 challenge.
We describe our solution for the PIRM Super-Resolution Challenge 2018 where we achieved the 2nd best perceptual quality for average RMSE<=16, 5th best for RMSE<=12.5, and 7th best for RMSE<=11.5. We modify a recently proposed Multi-Grid Back-Projection (MGBP) architecture to work as a generative system with an input parameter that can control the amount of artificial details in the output. We propose a discriminator for adversarial training with the following novel properties: it is multi-scale that resembles a progressive-GAN; it is recursive that balances the architecture of the generator; and it includes a new layer to capture significant statistics of natural images. Finally, we propose a training strategy that avoids conflicts between reconstruction and perceptual losses. Our configuration uses only 281k parameters and upscales each image of the competition in 0.2s in average.