DCUNet: Deformable Convolutional UNet for Multi-frame Infrared Small Target Super-resolution

The objective of infrared multi-frame super-resolution for small targets is to enhance the target's resolution by leveraging complementary information from multiple frames. However, the presence of motion variations and scale changes in infrared small targets introduces discontinuities in information between frames, posing challenges for super-resolution. To address the above challenges, we propose a multi-frame infrared small target super-resolution network based on deformable UNet. Specifically, the network takes multi-frame sequential infrared images as input and divides the input into reference frames and current frames. It utilizes deformable convolutions to align the frames to enhance the target features of the current frame. Additionally, the encoding structure of the UNet network is employed to extract multiscale features from the sequential images. By leveraging skip connections and a decoding structure, the network achieves the enhancement of target detail information and the fusion of multiscale features, ultimately outputting the super-resolved image. We conducted experiments on the IRDST and NUDT-MIRSDT datasets, and the results validate the practicality of our designed network.