Image Super-Resolution via Attention based Back Projection Networks

10 Oct 2019  ยท  Zhi-Song Liu, Li-Wen Wang, Chu-Tak Li, Wan-Chi Siu, Yui-Lam Chan ยท

Deep learning based image Super-Resolution (SR) has shown rapid development due to its ability of big data digestion. Generally, deeper and wider networks can extract richer feature maps and generate SR images with remarkable quality. However, the more complex network we have, the more time consumption is required for practical applications. It is important to have a simplified network for efficient image SR. In this paper, we propose an Attention based Back Projection Network (ABPN) for image super-resolution. Similar to some recent works, we believe that the back projection mechanism can be further developed for SR. Enhanced back projection blocks are suggested to iteratively update low- and high-resolution feature residues. Inspired by recent studies on attention models, we propose a Spatial Attention Block (SAB) to learn the cross-correlation across features at different layers. Based on the assumption that a good SR image should be close to the original LR image after down-sampling. We propose a Refined Back Projection Block (RBPB) for final reconstruction. Extensive experiments on some public and AIM2019 Image Super-Resolution Challenge datasets show that the proposed ABPN can provide state-of-the-art or even better performance in both quantitative and qualitative measurements.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Image Super-Resolution BSD100 - 16x upscaling ABPN PSNR 22.72 # 1
SSIM 0.512 # 1
Image Super-Resolution BSD100 - 4x upscaling ABPN PSNR 27.82 # 11
SSIM 0.743 # 18
Image Super-Resolution BSD100 - 8x upscaling ABPN PSNR 24.99 # 3
SSIM 0.604 # 3
Image Super-Resolution DIV2K val - 16x upscaling ABPN PSNR 24.38 # 1
SSIM 0.641 # 1
Image Super-Resolution DIV8K val - 16x upscaling ABPN PSNR 26.71 # 1
SSIM 0.65 # 1
Image Super-Resolution Manga109 - 16x upscaling ABPN PSNR 21.25 # 1
SSIM 0.673 # 1
Image Super-Resolution Manga109 - 4x upscaling ABPN PSNR 31.79 # 10
SSIM 0.921 # 14
Image Super-Resolution Manga109 - 8x upscaling ABPN PSNR 25.29 # 4
SSIM 0.802 # 4
Image Super-Resolution Set14 - 4x upscaling ABPN PSNR 28.94 # 21
SSIM 0.789 # 24
Image Super-Resolution Set14 - 8x upscaling ABPN PSNR 25.08 # 4
SSIM 0.638 # 5
Image Super-Resolution Set5 - 8x upscaling ABPN PSNR 27.25 # 5
SSIM 0.786 # 5
Image Super-Resolution Urban100 - 16x upscaling ABPN PSNR 20.39 # 1
SSIM 0.515 # 1
Image Super-Resolution Urban100 - 4x upscaling ABPN PSNR 27.06 # 12
SSIM 0.811 # 14
Image Super-Resolution Urban100 - 8x upscaling ABPN PSNR 23.04 # 4
SSIM 0.641 # 5

Methods


No methods listed for this paper. Add relevant methods here