Yukyung Choi* (KAIST), Namil Kim* (KAIST), Soonmin Hwang* (KAIST), Jongchan Park (KAIST), In So Kweon (KAIST)
The "SuperResolution benchmark" consists of 57 training image pairs and 10 test image pairs, saved in loss less png format. Since thermal images are provided in 16 bits format, we converted to 8 bits level to test and visualize the data. Note that we use a clipped linear mapping approaches for the contrast enhancement of thermal image as a pre-processing.
@INPROCEEDINGS{7759059,
author={Y. Choi and N. Kim and S. Hwang and I. S. Kweon},
booktitle={2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
title={Thermal Image Enhancement using Convolutional Neural Network},
year={2016},
pages={223-230},
month={Oct},}