FSIRR: Mining Frequency- and Spatial-Domain Features for Enhanced Infrared Image Recovery

Infrared (IR) images are widely used in various applications, with IR image restoration (IRR) serving as a crucial foundation for advanced tasks such as target detection. Despite deep learning’s dominance in IRR, the development has been hindered by the lack of publicly available datasets. To address this, we constructed a large-scale benchmark dataset comprising 2450 IR image deblurring pairs and 2000 IR image denoising pairs, incorporating diverse and complex backgrounds. In addition, we developed a benchmark framework for IRR, which uses spatial- and frequency-domain features to enhance IRR. Specifically, we design a spatial-domain residual feature extraction module (SDResM) to leverage critical spatial information and a frequency-domain-aware bottleneck (FDAB) to restore degraded IR images by using their frequency-domain features. Extensive experimental results show that the proposed method achieves state-of-the-art (SOTA) performance on the proposed dataset, enabling high-quality recovery of degraded IR images. The new IR denoising and deblurring dataset is available at https://github.com/aurora-sea/IRAY