Fused pyramid attention network for single image super‐resolution

Abstract

In image super‐resolution, deep neural networks with various attention mechanisms have achieved noticeable performance in recent years, for example, channel attention and layer attention. Although many researchers have achieved good super‐resolution results with only a certain style of attention, the divergence and the complementarity focused by multiple attention mechanisms are ignored. In addition, most of these methods fail to utilize the diverse information from multi‐scale features. To efficiently manipulate the above rich information, this paper strives to combine multi‐scale structure and multi‐attention schemes in architecture and module levels for super‐resolution. Especially, in the architecture level, a fused pyramid attention network is developed to extract deep features with the multi‐scale context information from multiple different sizes of receptive field recurrently with skip connections. For the module level, a fused pyramid attention module is designed to fuse the two attention mechanisms to further refine the deep features with fine‐grained information. Compared with the common fusion strategy, the adopted feature fusion structure can maintain better structural information while establishing long‐range dependency. Extensive experimental results demonstrate that the proposed network achieves favorable performance quantitatively and visually.

Super‐resolution reconstruction based on generative adversarial networks with dual branch half instance normalization

“…Chen et al. [38] proposed an attention in attention network including both non attention and attention branches with a dynamic attention module used to adaptively generate weights for both branches based on the input features. Xia et al.…”

Recursive Self-Attention Modules-Based Network for Panchromatic and Multispectral Image Fusion

“…We design a recursive self-attention module-based network (RSANet), which is a single branch network that is based on the concept of super-resolution [50]. We combine the PAN and the LRMS after up-sampling by a factor of 4 as the input to the network.…”

Lightweight frequency-based attention network for image super-resolution

“…Accordingly, many lightweight methods that incorporate attention mechanisms are constantly evolving. Hui et al 41 developed a contrast-aware CA mechanism at the end of the progressive refinement module to steadily boost the accuracy of image SR. Chen et al 42 proposed attention in an attention network, in which a non-attention branch and a coupling attention branch adjust the contribution of attention layers dynamically. Wan et al 43 built a hierarchical SA mechanism to jointly and adaptively fuse local and global hierarchical features for HR reconstruction.…”

“…Hui et al 41 . developed a contrast-aware CA mechanism at the end of the progressive refinement module to steadily boost the accuracy of image SR. Chen et al 42 . proposed attention in an attention network, in which a non-attention branch and a coupling attention branch adjust the contribution of attention layers dynamically.…”