PLSMS model for restoration of the details concealed by light sources in nighttime hazed image
- PDF / 4,113,740 Bytes
- 9 Pages / 595.276 x 790.866 pts Page_size
- 4 Downloads / 183 Views
ORIGINAL PAPER
PLSMS model for restoration of the details concealed by light sources in nighttime hazed image Chunming Tang1 · Ruiyu Sun2 · Zheng Lian2 · Wenyan Zhu2 Received: 2 July 2019 / Revised: 27 February 2020 / Accepted: 3 August 2020 © Springer-Verlag London Ltd., part of Springer Nature 2020
Abstract In the popular atmospheric scattering model applied to image dehazed, the scatter from the adjacent light sources on the camera sensor which has degraded the image quality is neglected, especially when some artificial light sources exist. Therefore, in the dehazing results, the light sources are always much brighter than those in the original image in order to enhance the majority details in the dark regions, which causes the details around the light sources to be obscured. We propose a novel image restoration model mainly based on point light source multiple scattering theory. An observed hazed image can be described as a linear combination of light being reflected from an imaged object itself and its multiple scattering component. The restoration quality of the artificial light sources and details around them may be improved by suppressing the multiple scattering which can be simulated by an APSF function. The key of our model is to find the appropriate APSF kernels by analysis and some experiments. Comparisons and evaluations of our restoration results with those from popular algorithms show that ours is better in details and colors recovery in dealing with light sources in nighttime hazed images. Keywords Image restoration · Restoration of the details concealed by light sources · Point light source multiple scattering model · Atmospheric point spread function
1 Introduction Captured images are often affected by the severe weather and other unfavorable conditions, such as light attenuation by fog and low illumination at night, resulting in degradation. Research on image restoration seeks out practical solutions to these challenging and widespread problems. In general, fog or hazed weather effects more on captured images because of more severe light attenuation and * Chunming Tang [email protected] Ruiyu Sun [email protected] Zheng Lian [email protected] Wenyan Zhu [email protected] 1
Tianjin Key Laboratory of Photoelectric Detection Technology and System, School of Artificial Intelligent, Tianjin Polytechnic University, Tianjin 300387, China
School of Electronics and Information Engineering, Tianjin Polytechnic University, Tianjin 300387, China
2
scattering as more particles in the air. Many single imagebased dehazing algorithms have been implemented via Middleton’s model [1–19] so far, based on some prior image information, such as dark channel [1, 20, 21], color attenuation [14] and hazed line [3], to estimate its two unknowns, namely the transmission map and the airlight, assumed to be constant. However, this model works better in daytime than nighttime. At night, the airlight is replaced by the strongly non-uniform and varicolored multiple colored artificial light sources, such as
Data Loading...
