Multi-Scale Ensemble Learning for Thermal Image Enhancement

In this study, we propose a multi-scale ensemble learning method for thermal image enhancement in different scale conditions based on convolutional neural networks. Incorporating the multiple scales of images has been tricky task so that methods have individually trained and evaluated each scale. However, leads to limitation network properly operates specific To address issue, novel parallel architecture leveraging confidence maps introduced train well varying conditions. The experimental results show our proposed outperforms conventional methods. evaluation is presented both quantitatively qualitatively.