In this supplemental material, we first provide additional ablation experiments on the activation function, the normalization layer and the introduced frequency loss in Sec. 1. Our ShuffleMixercan be applied toother restoration tasks. Table 1: Ablation experiments forcomponents oftheShuffleMixerLayer and the frequencyloss function. We evaluate them respectively on the developed tiny model and train them on the 4 DIV2Kdataset. Why does ShuffleMixer perform poorly on PSNR?This result may be caused by the luminance differences in some local areas.

This is a classical problem that has attracted lots of attention recently due to the rapid developmentofhigh-definition devices,suchasUltra-High Definition Television,Samsung GalaxyS22 Ultra,iPhone13ProMax,andHUAWEIP50Pro,andsoon.

Lightweight and efficiency are critical drivers for the practical application of image super-resolution (SR) algorithms. We propose a simple and effective approach, ShuffleMixer, for lightweight image super-resolution that explores large convolution and channel split-shuffle operation. In contrast to previous SR models that simply stack multiple small kernel convolutions or complex operators to learn representations, we explore a large kernel ConvNet for mobile-friendly SR design. Specifically, we develop a large depth-wise convolution and two projection layers based on channel splitting and shuffling as the basic component to mix features efficiently. Since the contexts of natural images are strongly locally correlated, using large depth-wise convolutions only is insufficient to reconstruct fine details. To overcome this problem while maintaining the efficiency of the proposed module, we introduce Fused-MBConvs into the proposed network to model the local connectivity of different features. Experimental results demonstrate that the proposed ShuffleMixer is about $3 \times$ smaller than the state-of-the-art efficient SR methods, e.g.

Lightweight and efficiency are critical drivers for the practical application of image super-resolution (SR) algorithms. We propose a simple and effective approach, ShuffleMixer, for lightweight image super-resolution that explores large convolution and channel split-shuffle operation. In contrast to previous SR models that simply stack multiple small kernel convolutions or complex operators to learn representations, we explore a large kernel ConvNet for mobile-friendly SR design. Specifically, we develop a large depth-wise convolution and two projection layers based on channel splitting and shuffling as the basic component to mix features efficiently. Since the contexts of natural images are strongly locally correlated, using large depth-wise convolutions only is insufficient to reconstruct fine details. To overcome this problem while maintaining the efficiency of the proposed module, we introduce Fused-MBConvs into the proposed network to model the local connectivity of different features.