Effective estimation of background light in underwater image dehazing
Miao Yang, Bo Yin, Zhiqiang Wei, Yixiang Du, Jintong Hu · OSA Continuum · 2019
Background estimation is a crucial step in underwater image dehazing. Most of the current estimation methods assume a uniform background light in the underwater environment and select the brightest pixel in the dark channel as the candidate, which fails to explain the real interactions of light rays and particles in the water medium and causes over-saturation in dehazed images. In this paper, the relationship between the maxima of dark channel and the background light in offshore underwater images is initially illustrated, and a contradiction of the assumption related to the dark channel prior used in underwater image restoration is addressed. To the best of our knowledge, this is the first work studying the statistical facts of underwater background light. Furthermore, a machine learning based background light estimation and reconstruction method is proposed based on the learning of the maximum areas of a dark channel. The subjective and objective restoration results of the state-of-the-art algorithms with and without applying the proposed background light estimation method to the offshore images are compared. The results show that the proposed method better simulated the directional distribution of the background light in a turbid water environment, and the foggy ambiguity caused by the backscattering was removed more efficiently in comparison with existing methods.