Polarimetric underwater image recovery via deep learning. (October 2020)

• Record Type: Journal Article
• Authors: Hu, Haofeng; Zhang, Yanbin; Li, Xiaobo; Lin, Yang; Cheng, Zhenzhou; Liu, Tiegen
• Is Part Of: Optics and lasers in engineering. Volume 133(2020)
• Publication Date: 2020-10
• Subjects: Polarization -- Polarimetric imaging -- Image Recovery -- Deep Learning -- Dense network; Lasers in engineering -- Periodicals; Optical measurements -- Periodicals; Optics -- Periodicals; Lasers en ingénierie -- Périodiques; Mesures optiques -- Périodiques; Optique -- Périodiques; 621.36605
• DOI: 10.1016/j.optlaseng.2020.106152
• Languages: English
• ISSNs: 0143-8166

Abstract

Highlights: This article proposes, for the first time to our knowledge, a learning-based method for underwater polarimetric image recovery. The comparison experiment with the intensity or polarization information input demonstrates that additionally introducing the polarization information is beneficial for improving the image quality. The experimental results for different scenes in dense turbid water demonstrate that our learning-based method can effectively remove the veiling light to realize the clear vision and outperforms other existing methods. Abstract: Polarimetric imaging is an effective way for clear vision in water, in which deducing the object radiance in clear water from the obtained polarimetric information in turbid water is essential. In this letter, we propose, for the first time to our knowledge, a learning-based method for polarimetric underwater image recovery. It is based on the dense network and can learn well the relation between the object radiance and the polarization information. The experimental results demonstrate that additionally introducing the polarization information is beneficial for improving the image quality. Moreover, the proposed learning-based method can effectively remove the veiling light and outperforms other existing methods, even in dense turbid water.