Feedback Network for Image Super-Resolution

Abstract

The rapid development of deep learning (DL) has driven single image super-resolution (SR) into a new era. However, in most existing DL based image SR networks, the information flows are solely feedforward, and the high-level features cannot be fully explored. In this paper, we propose the gated multiple feedback network (GMFN) for accurate image SR, in which the representation of low-level features are efficiently enriched by rerouting multiple high-level features. We cascade multiple residual dense blocks (RDBs) and recurrently unfolds them across time. The multiple feedback connections between two adjacent time steps in the proposed GMFN exploits multiple high-level features captured under large receptive fields to refine the low-level features lacking enough contextual information. The elaborately designed gated feedback module (GFM) efficiently selects and further enhances useful information from multiple rerouted high-level features, and then refine the low-level features with the enhanced high-level information. Extensive experiments demonstrate the superiority of our proposed GMFN against stateof-the-art SR methods in terms of both quantitative metrics and visual quality. Code is available at https://github.com/liqilei/GMFN.

Lightweight Image Super-Resolution with Information Multi-distillation Network

“…Similarly, Zhang et al [37] utilized non-local attention to better guide feature extraction in their trunk branch for reaching better performance. Very recently, Li et al [17] exploited feedback mechanism that enhancing low-level representation with high-level ones. For lightweight networks, Hui et al [11] developed the information distillation network for better exploiting hierarchical features by separation processing of the current feature maps.…”

Versatile Video Coding and Super-Resolution for Efficient Delivery of 8k Video with 4k Backward-Compatibility

“…The baseline version of SRFBN provided by the author in [17] is trained to recover the high-resolution version of uncompressed lowresolution data. In our case, we focus on assessing this method on video presenting compression artefacts.…”

“…Although VMAF is optimized for visual quality estimation of 4K contents, it can be relevant to add this evaluation method in the experiment as it proposes a high correlation with subjective test ratings. For the pre/post-processing pipeline, we use the deep-learning-based method Super-Resolution with Feedback Network (SRFBN) [17], which enables good performance in both visual quality improvement and runtime. We have trained the model using compressed data to propose a fair evaluation of this method on contents presenting compression artefacts.…”

Adaptive Densely Connected Single Image Super-Resolution

“…And it employs lightweight networks to achieve fast and accurate results. The SR feedback network (SRFBN) [15] proposed by Li et al applies the RNN with constraints to process feedback information and performs feature reuse.…”