Dan Zhu

Pablo Navarrete Michelini, Wenbin Chen, Hanwen Liu et al.

Multi-Grid Back-Projection (MGBP) is a fully-convolutional network architecture that can learn to restore images and videos with upscaling artifacts. Using the same strategy of multi-grid partial differential equation (PDE) solvers this multiscale architecture scales computational complexity efficiently with increasing output resolutions. The basic processing block is inspired in the iterative back-projection (IBP) algorithm and constitutes a type of cross-scale residual block with feedback from low resolution references. The architecture performs in par with state-of-the-arts alternatives for regression targets that aim to recover an exact copy of a high resolution image or video from which only a downscale image is known. A perceptual quality target aims to create more realistic outputs by introducing artificial changes that can be different from a high resolution original content as long as they are consistent with the low resolution input. For this target we propose a strategy using noise inputs in different resolution scales to control the amount of artificial details generated in the output. The noise input controls the amount of innovation that the network uses to create artificial realistic details. The effectiveness of this strategy is shown in benchmarks and it is explained as a particular strategy to traverse the perception-distortion plane.

Pablo Navarrete Michelini, Dan Zhu, Hanwen Liu

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.