Fully 1 × 1 Convolutional Network for Lightweight Image Super-resolution

Deep convolutional neural networks, particularly large models with large kernels (3\times3 or more), have achieved significant progress in single image super-resolution (SISR) tasks. However, the heavy computational footprint of such models prevents their deployment in real-time, resource-constrained environments. Conversely, 1\times1 convolutions have substantial computational efficiency, but struggle with aggregating local spatial representations, which is an essential capability for SISR models. In response to this dichotomy, we propose to harmonize the merits of both 3\times3 and 1\times1 kernels, and exploit their great potential for lightweight SISR tasks. Specifically, we propose a simple yet effective fully 1\times1 convolutional network, named shift-Conv-based network (SCNet). By incorporating a parameter-free spatial-shift operation, the fully 1\times1 convolutional network is equipped with a powerful representation capability and impressive computational efficiency. Extensive experiments demonstrate that SCNets, despite their fully 1\times1 convolutional structure, consistently match or even surpass the performance of existing lightweight SR models that employ regular convolutions. The code and pretrained models can be found at https://github.com/Aitical/SCNet.