TY - GEN
T1 - An introductory study on image quality of dehazed images
AU - Azizah, Aruni Yasmin
AU - Rahadianti, Laksmita
AU - Deborah, Hilda
N1 - Funding Information:
The authors would like to thank University Inodonesia, for funding us through Hibah Publikasi Terindeks International (PUTI) Q3 No. NKB-1822/UN2.RST/HKP.05.00/2020. Additionally, the authors would like to thank the Tokopedia-UI AI Center of Excellence of Fasilkom UI for providing the infrastructure of NVIDIA DGX-1 for traing our deep learing models. H. Deborah collaboration and work visit is supported by FRIPRO FRINATEK Metrological texture analysis for hyperspectral images (projectnr. 274881) funded by the Research Council of Norway.
Publisher Copyright:
© 2020 IEEE.
PY - 2020/10/17
Y1 - 2020/10/17
N2 - In real World conditions, e.g., fog and haze, the surrounding environment contains a large number of microparticles. This creates a medium which interferes with propagating light, changing the appearance of objects within it. When a digital image is captured in this type of medium, the image will be hazy due to blurring and loss of detail. This is a problem for many automated computer vision methods which utilize visual cues in the image. Multiple methods have been proposed to perform image dehazing, but it is difficult to standardize the perceived quality of the estimated clear image. A highly accurate estimation is not always possible because the nature of the problem is m-posed and an exact clear ground truth is usually unavailable. In fact, the goal of many dehazing methods is to simply obtain a visually pleasing understandable image. In this paper, the Pix2pix image to image translation network and the Contrast Limited Adaptive Histogram Equalization (CLAHE) image enhancement method are used to dehaze hazy images. The resulting clear images will be subject to both subjective and objective image quality assessments. Lastly, this paper will also provide an analysis to discuss the important factors in the attempt to obtain a better dehazed image.
AB - In real World conditions, e.g., fog and haze, the surrounding environment contains a large number of microparticles. This creates a medium which interferes with propagating light, changing the appearance of objects within it. When a digital image is captured in this type of medium, the image will be hazy due to blurring and loss of detail. This is a problem for many automated computer vision methods which utilize visual cues in the image. Multiple methods have been proposed to perform image dehazing, but it is difficult to standardize the perceived quality of the estimated clear image. A highly accurate estimation is not always possible because the nature of the problem is m-posed and an exact clear ground truth is usually unavailable. In fact, the goal of many dehazing methods is to simply obtain a visually pleasing understandable image. In this paper, the Pix2pix image to image translation network and the Contrast Limited Adaptive Histogram Equalization (CLAHE) image enhancement method are used to dehaze hazy images. The resulting clear images will be subject to both subjective and objective image quality assessments. Lastly, this paper will also provide an analysis to discuss the important factors in the attempt to obtain a better dehazed image.
UR - http://www.scopus.com/inward/record.url?scp=85099748303&partnerID=8YFLogxK
U2 - 10.1109/ICACSIS51025.2020.9263131
DO - 10.1109/ICACSIS51025.2020.9263131
M3 - Conference contribution
AN - SCOPUS:85099748303
T3 - 2020 International Conference on Advanced Computer Science and Information Systems, ICACSIS 2020
SP - 301
EP - 308
BT - 2020 International Conference on Advanced Computer Science and Information Systems, ICACSIS 2020
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 12th International Conference on Advanced Computer Science and Information Systems, ICACSIS 2020
Y2 - 17 October 2020 through 18 October 2020
ER -