Brightness-based convolutional neural network for thermal image enhancement

Kyungjae Lee, Junhyeop Lee, Joosung Lee, Sangwon Hwang, Sangyoun Lee

Research output: Contribution to journalArticle

13 Citations (Scopus)

Abstract

In this paper, we propose a convolutional neural network for thermal image enhancement by incorporating the brightness domain with a residual-learning technique, which improves the performance of enhancement and speed of convergence. Typically, the training domain uses the same domain as that of the target image; however, we evaluated several domains to determine the most suitable one for the network. In the analyses, we first compared the performance of networks that were trained by the corresponding regions of color-based and aligned infrared-based images, respectively, including thermal, far, and near spectra. Then, four RGB-based domains, namely, gray, lightness, intensity, and brightness were evaluated. Finally, the proposed network architecture was determined by considering the residual and brightness domains. The results of the analyses indicated that the brightness domain was the best training domain for enhancing the thermal images. The experimental results confirm that the proposed network, which can be trained in approximately one hour, outperforms the conventional learning-based approaches for thermal image enhancement, in terms of several image quality metrics and a qualitative evaluation. Furthermore, the results demonstrate that the brightness domain is effective as the training domain and can be used to increase the performance of existing networks.

Original languageEnglish
Article number8094863
Pages (from-to)26867-26879
Number of pages13
JournalIEEE Access
Volume5
DOIs
Publication statusPublished - 2017 Nov 2

Fingerprint

Image enhancement
Luminance
Neural networks
Network architecture
Image quality
Hot Temperature
Color
Infrared radiation

All Science Journal Classification (ASJC) codes

  • Computer Science(all)
  • Materials Science(all)
  • Engineering(all)

Cite this

Lee, Kyungjae ; Lee, Junhyeop ; Lee, Joosung ; Hwang, Sangwon ; Lee, Sangyoun. / Brightness-based convolutional neural network for thermal image enhancement. In: IEEE Access. 2017 ; Vol. 5. pp. 26867-26879.
@article{22c62adb25bf412898114d0bae77edc0,
title = "Brightness-based convolutional neural network for thermal image enhancement",
abstract = "In this paper, we propose a convolutional neural network for thermal image enhancement by incorporating the brightness domain with a residual-learning technique, which improves the performance of enhancement and speed of convergence. Typically, the training domain uses the same domain as that of the target image; however, we evaluated several domains to determine the most suitable one for the network. In the analyses, we first compared the performance of networks that were trained by the corresponding regions of color-based and aligned infrared-based images, respectively, including thermal, far, and near spectra. Then, four RGB-based domains, namely, gray, lightness, intensity, and brightness were evaluated. Finally, the proposed network architecture was determined by considering the residual and brightness domains. The results of the analyses indicated that the brightness domain was the best training domain for enhancing the thermal images. The experimental results confirm that the proposed network, which can be trained in approximately one hour, outperforms the conventional learning-based approaches for thermal image enhancement, in terms of several image quality metrics and a qualitative evaluation. Furthermore, the results demonstrate that the brightness domain is effective as the training domain and can be used to increase the performance of existing networks.",
author = "Kyungjae Lee and Junhyeop Lee and Joosung Lee and Sangwon Hwang and Sangyoun Lee",
year = "2017",
month = "11",
day = "2",
doi = "10.1109/ACCESS.2017.2769687",
language = "English",
volume = "5",
pages = "26867--26879",
journal = "IEEE Access",
issn = "2169-3536",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

Brightness-based convolutional neural network for thermal image enhancement. / Lee, Kyungjae; Lee, Junhyeop; Lee, Joosung; Hwang, Sangwon; Lee, Sangyoun.

In: IEEE Access, Vol. 5, 8094863, 02.11.2017, p. 26867-26879.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Brightness-based convolutional neural network for thermal image enhancement

AU - Lee, Kyungjae

AU - Lee, Junhyeop

AU - Lee, Joosung

AU - Hwang, Sangwon

AU - Lee, Sangyoun

PY - 2017/11/2

Y1 - 2017/11/2

N2 - In this paper, we propose a convolutional neural network for thermal image enhancement by incorporating the brightness domain with a residual-learning technique, which improves the performance of enhancement and speed of convergence. Typically, the training domain uses the same domain as that of the target image; however, we evaluated several domains to determine the most suitable one for the network. In the analyses, we first compared the performance of networks that were trained by the corresponding regions of color-based and aligned infrared-based images, respectively, including thermal, far, and near spectra. Then, four RGB-based domains, namely, gray, lightness, intensity, and brightness were evaluated. Finally, the proposed network architecture was determined by considering the residual and brightness domains. The results of the analyses indicated that the brightness domain was the best training domain for enhancing the thermal images. The experimental results confirm that the proposed network, which can be trained in approximately one hour, outperforms the conventional learning-based approaches for thermal image enhancement, in terms of several image quality metrics and a qualitative evaluation. Furthermore, the results demonstrate that the brightness domain is effective as the training domain and can be used to increase the performance of existing networks.

AB - In this paper, we propose a convolutional neural network for thermal image enhancement by incorporating the brightness domain with a residual-learning technique, which improves the performance of enhancement and speed of convergence. Typically, the training domain uses the same domain as that of the target image; however, we evaluated several domains to determine the most suitable one for the network. In the analyses, we first compared the performance of networks that were trained by the corresponding regions of color-based and aligned infrared-based images, respectively, including thermal, far, and near spectra. Then, four RGB-based domains, namely, gray, lightness, intensity, and brightness were evaluated. Finally, the proposed network architecture was determined by considering the residual and brightness domains. The results of the analyses indicated that the brightness domain was the best training domain for enhancing the thermal images. The experimental results confirm that the proposed network, which can be trained in approximately one hour, outperforms the conventional learning-based approaches for thermal image enhancement, in terms of several image quality metrics and a qualitative evaluation. Furthermore, the results demonstrate that the brightness domain is effective as the training domain and can be used to increase the performance of existing networks.

UR - http://www.scopus.com/inward/record.url?scp=85033698596&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85033698596&partnerID=8YFLogxK

U2 - 10.1109/ACCESS.2017.2769687

DO - 10.1109/ACCESS.2017.2769687

M3 - Article

AN - SCOPUS:85033698596

VL - 5

SP - 26867

EP - 26879

JO - IEEE Access

JF - IEEE Access

SN - 2169-3536

M1 - 8094863

ER -