Then, we give an illustration of operation details in the GraphAgg. B presents further analysis and discussions onour proposed GraphAgg module and IGNN network. Denote the feature shapes ofEL s and EL as H/s W/s and H W respectively. Each LR patch ofEL find thek nearest neighboring LR patches fromEL s. In this section, we first present more ablation experiments to demonstrate the effectiveness of the proposedIGNNfurther,includingtheeffectofusing F0LandFL sandnumberofGraphAggmodules insertedinnetworks.

We provide two variant models for image SR, called CA T -R-2 and CA T -A-2. The MLP expansion ratio is set as 4. We use self-ensemble strategy and mark models with "+". The results are shown in Table 1. Output size is 3 512 512 to calculate FLOPs. CA T -R-2 achieves 0.22 dB on Urban100 ( All these results further indicate the effectiveness of our method.

Our proposed aligned structured sparsity learning (ASSL) algorithm is summarized in Algorithm 1. There are in total 16 residual blocks in EDSR_baseline. We provide more visual comparisons in Figure 1. In contrast, our ASSLN can better recover more structural details. While, our ASSLN can better alleviate the blurring artifacts.

Our proposed aligned structured sparsity learning (ASSL) algorithm is summarized in Algorithm 1. There are in total 16 residual blocks in EDSR_baseline. We provide more visual comparisons in Figure 1. In contrast, our ASSLN can better recover more structural details. While, our ASSLN can better alleviate the blurring artifacts.

In this paper, we explore the cross-scale patch recurrence property of a natural image, i.e.,

We provide two variant models for image SR, called CA T -R-2 and CA T -A-2. The MLP expansion ratio is set as 4. We use self-ensemble strategy and mark models with "+". The results are shown in Table 1. Output size is 3 512 512 to calculate FLOPs. CA T -R-2 achieves 0.22 dB on Urban100 ( All these results further indicate the effectiveness of our method.

Transformer-based Super-Resolution (SR) models have recently advanced image reconstruction quality, yet challenges remain due to computational complexity and an over-reliance on large patch sizes, which constrain fine-grained detail enhancement. In this work, we propose TaylorIR to address these limitations by utilizing a patch size of 1x1, enabling pixel-level processing in any transformer-based SR model. To address the significant computational demands under the traditional self-attention mechanism, we employ the TaylorShift attention mechanism, a memory-efficient alternative based on Taylor series expansion, achieving full token-to-token interactions with linear complexity. Experimental results demonstrate that our approach achieves new state-of-the-art SR performance while reducing memory consumption by up to 60% compared to traditional self-attention-based transformers.

Self-similarity refers to the image prior widely used in image restoration algorithms that small but similar patterns tend to occur at different locations and scales. However, recent advanced deep convolutional neural network based methods for image restoration do not take full advantage of self-similarities by relying on self-attention neural modules that only process information at the same scale. To solve this problem, we present a novel Pyramid Attention module for image restoration, which captures long-range feature correspondences from a multi-scale feature pyramid. Inspired by the fact that corruptions, such as noise or compression artifacts, drop drastically at coarser image scales, our attention module is designed to be able to borrow clean signals from their "clean" correspondences at the coarser levels. The proposed pyramid attention module is a generic building block that can be flexibly integrated into various neural architectures. Its effectiveness is validated through extensive experiments on multiple image restoration tasks: image denoising, demosaicing, compression artifact reduction, and super resolution. Without any bells and whistles, our PANet (pyramid attention module with simple network backbones) can produce state-of-the-art results with superior accuracy and visual quality. Our code will be available at https://github.com/SHI-Labs/Pyramid-Attention-Networks