Non-Local Recurrent Network for Image Restoration

Many classic methods have shown non-local self-similarity in natural images to be an effective prior for image restoration. However, it remains unclear and challenging to make use of this intrinsic property via deep networks. In this paper, we propose a non-local recurrent network (NLRN) as the first attempt to incorporate non-local operations into a recurrent neural network (RNN) for image restoration. The main contributions of this work are: (1) Unlike existing methods that measure self-similarity in an isolated manner, the proposed non-local module can be flexibly integrated into existing deep networks for end-to-end training to capture deep feature correlation between each location and its neighborhood. (2) We fully employ the RNN structure for its parameter efficiency and allow deep feature correlation to be propagated along adjacent recurrent states. This new design boosts robustness against inaccurate correlation estimation due to severely degraded images. (3) We show that it is essential to maintain a confined neighborhood for computing deep feature correlation given degraded images. This is in contrast to existing practice that deploys the whole image. Extensive experiments on both image denoising and super-resolution tasks are conducted. Thanks to the recurrent non-local operations and correlation propagation, the proposed NLRN achieves superior results to state-of-the-art methods with much fewer parameters.

PDF Abstract NeurIPS 2018 PDF NeurIPS 2018 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Image Super-Resolution BSD100 - 4x upscaling NLRN PSNR 27.48 # 32
SSIM 0.7306 # 35
Grayscale Image Denoising BSD200 sigma30 NLRN-MV PSNR 28.2 # 2
Grayscale Image Denoising BSD200 sigma50 NLRN-MV PSNR 25.97 # 2
Grayscale Image Denoising BSD200 sigma70 NLRN-MV PSNR 24.62 # 2
Grayscale Image Denoising BSD68 sigma15 NLRN PSNR 31.88 # 5
Grayscale Image Denoising BSD68 sigma25 NLRN PSNR 29.41 # 3
Grayscale Image Denoising BSD68 sigma50 NLRN PSNR 26.47 # 5
Denoising Darmstadt Noise Dataset NLRN PSNR 30.8 # 10
Grayscale Image Denoising Set12 sigma15 NLRN PSNR 33.16 # 2
Grayscale Image Denoising Set12 sigma30 NLRN PSNR 30.8 # 1
Grayscale Image Denoising Set12 sigma50 NLRN PSNR 27.64 # 3
Image Super-Resolution Set14 - 4x upscaling NLRN PSNR 28.36 # 50
SSIM 0.7745 # 52
Image Super-Resolution Urban100 - 4x upscaling NLRN PSNR 25.79 # 35
SSIM 0.7729 # 33
Grayscale Image Denoising Urban100 sigma15 NLRN PSNR 33.45 # 4
Grayscale Image Denoising Urban100 sigma25 NLRN PSNR 30.94 # 6
Grayscale Image Denoising Urban100 sigma50 NLRN PSNR 27.49 # 5


No methods listed for this paper. Add relevant methods here