Fieldscale: Locality-Aware Field-based Adaptive Rescaling for Thermal Infrared Image

Hyeonjae Gil¹ · Myung-Hwan Jeon¹ · Ayoung Kim^1*

¹Seoul National University

RA-L 2024 (and ICRA 2025)

[Paper] [Supplementary Video] [BibTex]

What is Fieldscale?

Fieldscale is a TIR image rescaling method that adaptively rescales TIR images by considering both the intensity value and spatial context of each pixel. Output TIR images exhibit enhanced image quality and strong usability for downstream tasks such as object detection and place recognition.

How does it work?

Fieldscale constructs two 2D scalar fields, the min field and the max field, to encode the local intensity range of each pixel. The fields are then used to adaptively rescale the input image. Although Fieldscale is designed for TIR images, it can also be applied to the videos.

Comparison with Existing Methods

Fieldscale effectively rescales TIR images while preserving the local details and enhancing the global consistency. We note that FLIR AGC is only provided by the FLIR thermal camera manufacturer and Fieldscale can be a general-purpose method for various TIR cameras.

How to use

Fieldscale is currently implemented in Python (C++ implementation will be available soon). Please refer to the python directory for more details.

News

• 2025-01-28: Fieldscale will be presented at the IEEE International Conference on Robotics and Automation (ICRA) 2025.
• 2024-06-07: All code are now available on GitHub.
• 2024-05-17: Fieldscale has been accepted to IEEE Robotics and Automation Letters (RA-L)!

Citing Fieldscale

If you find this repository useful, please consider giving a star ⭐ and citing:

@article{gil2024fieldscale,
  title={Fieldscale: Locality-Aware Field-based Adaptive Rescaling for Thermal Infrared Image},
  author={Gil, Hyeonjae and Jeon, Myung-Hwan and Kim, Ayoung},
  journal={IEEE Robotics and Automation Letters},
  year={2024},
  publisher={IEEE}
}