Haze removal algorithm using improved restoration model based on dark channel prior / Dai Zhen

• Main Author: Dai , Zhen
• Format: Thesis
• Published: 2019
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://studentsrepo.um.edu.my/10751/1/Dai_Zhen.pdf; http://studentsrepo.um.edu.my/10751/2/Dai_Zhen_%E2%80%93_Dissertation.pdf; http://studentsrepo.um.edu.my/10751/

In recent years, with the rapid development of social economy and the continuous improvement of people’s living standard, the awareness of security precaution is becoming increasingly important. As an important tool for security work, video monitoring is widely used in traffic, outdoors, shopping malls and warehouses. However, in severe weather conditions, rain and haze have a large influence on the images obtained by video monitoring like the image contrast declines, color fades and edge blur. In this way, it is difficult to obtain the image contrast information, and it has a bad impact on the security work. So the clarity of images becomes very meaningful, and researchers start to pay attention to the field of image dehazing. Among many studies, the dark channel prior (DCP) dehazing algorithm is a major breakthrough in the field of image dehazing technology, and this algorithm has the advantages of simple, real-time and effective, while the limitation is that its shortcomings are mainly on the atmospheric scatting physical model over-reliance, and cannot be chosen to match the size of the filter temple, the lack of applicability of the sky and the image after the dark side. Based on the atmospheric physical model, this research proposed a haze removal algorithm using improved restoration model based on dark channel prior, which has good robustness to the bright sky regions, and also has a good effect on the edges. Firstly, the haze images were detected as having sky regions or having no sky regions. Secondly, haze images that have sky regions were segmented into sky region and non-sky region. Then, the haze images that have sky regions and have no sky regions were dehazed respectively. Finally, the haze-free images were obtained. For the haze-free images obtained in this research, the subjective and objective evaluation criteria are adopted. Subjective evaluation is mainly in accordance with human visual standards, and the common objective evaluation criteria using “Image Quality Evaluator (NIQE)” and also compare the proposed result with other traditional methods.