Distributed Kalman Filtering for Temporal Resolution Enhancement of Thermal Images

This paper presents a novel method for enhancing temporal resolution of thermal image sequences. The proposed approach employs the Kalman filter, which exploits the spatio-temporal correlation of the underlying physical processes and leverages temporal information contained in rows of an image captured via rolling shutter readout, to successfully reconstruct intra-frame sequences. To address the computational complexity associated with large-scale Kalman filtering, we propose a distributed filtering scheme that takes advantage of the neighbourhood relationship in the underlying diffusion process. Simulation results demonstrate the capability of the method in enhancing temporal resolution while achieving acceptable Peak Signal-to-Noise Ratio (PSNR) of approximately
$30dB$
. The distributed Kalman filter offers significant reductions in computational and storage costs, making it highly scalable for large images.