Thermal Video Enhancement Mamba: A Novel Approach to Thermal Video Enhancement for Real-World Applications

Information, Journal Year: 2025, Volume and Issue: 16(2), P. 125 - 125

Published: Feb. 9, 2025

Object tracking in thermal video is challenging due to noise, blur, and low contrast. We present TVEMamba, a Mamba-based enhancement framework with near-linear complexity that improves these conditions. Our approach uses State Space 2D (SS2D) module integrated Convolutional Neural Networks (CNNs) filter, sharpen, highlight important details. Key components include (i) denoising reduce background noise enhance image clarity, (ii) an optical flow attention handle complex motion (iii) entropy-based labeling create fully labeled dataset for training evaluation. TVEMamba outperforms existing methods (DCRGC, RLBHE, IE-CGAN, BBCNN) across multiple datasets (BIRDSAI, FLIR, CAMEL, Autonomous Vehicles, Solar Panels) achieves higher scores on standard quality metrics (EME, BDIM, DMTE, MDIMTE, LGTA). Extensive tests, including ablation studies convergence analysis, confirm its robustness. Real-world examples, such as humans, animals, moving objects self-driving vehicles remote sensing, demonstrate the practical value of TVEMamba.

Language: Английский

Neural Memory Self-Supervised State Space Models With Learnable Gates

IEEE Signal Processing Letters, Journal Year: 2025, Volume and Issue: 32, P. 926 - 930

Published: Jan. 1, 2025

Language: Английский