Enhancing Object Detection in Low Quality Images Using Deep Neural Network Based Super-Resolution 


Vol. 45,  No. 12, pp. 2169-2176, Dec.  2020
10.7840/kics.2020.45.12.2169


PDF Full-Text
  Abstract

Obtaining potential gains from object detection in practical systems hinges on sufficient levels of image resolution. In this paper, a two-step guideline using deep neural network (DNN)-based super-resolution (SR) model for object detection is presented under the assumption that the DNN model takes multiple low-resolution images captured form the same scene. In the first step, enhanced deep residual networks for single image super-resolution (EDSR) is exploited to recover an intermediate high-resolution image. The second step is to perform face and object detection based on RetinaFace and EfficientDet-D7. By capturing different training and test image formats, the resulting design employs a transfer learning method with pre-trained EDSR to further leverage detection performance. We adopt three degradation models for performance analysis and provide a practical guideline for DNN-based SR reconstruction. Numerical results show the effectiveness of the proposed method in improving small-sized object detection performance.

  Statistics
Cumulative Counts from November, 2022
Multiple requests among the same browser session are counted as one view. If you mouse over a chart, the values of data points will be shown.


  Cite this article

[IEEE Style]

D. Kim, Y. Yoo, S. Noh, J. Lee, "Enhancing Object Detection in Low Quality Images Using Deep Neural Network Based Super-Resolution," The Journal of Korean Institute of Communications and Information Sciences, vol. 45, no. 12, pp. 2169-2176, 2020. DOI: 10.7840/kics.2020.45.12.2169.

[ACM Style]

Daehee Kim, Youngjun Yoo, Song Noh, and Jaekoo Lee. 2020. Enhancing Object Detection in Low Quality Images Using Deep Neural Network Based Super-Resolution. The Journal of Korean Institute of Communications and Information Sciences, 45, 12, (2020), 2169-2176. DOI: 10.7840/kics.2020.45.12.2169.

[KICS Style]

Daehee Kim, Youngjun Yoo, Song Noh, Jaekoo Lee, "Enhancing Object Detection in Low Quality Images Using Deep Neural Network Based Super-Resolution," The Journal of Korean Institute of Communications and Information Sciences, vol. 45, no. 12, pp. 2169-2176, 12. 2020. (https://doi.org/10.7840/kics.2020.45.12.2169)

Search
(IN TITLE, AUTHOR, ABSTRACT,KEYWORDS)

Advanced Search ▼

POPULAR KEYWORDS
(TOP 10 KEYWORDS)

Recent Publications
(LAST 3 YEARS)






pISSN : 1226 - 4717
eISSN : 2287 - 3880