Inshore eutrophic underwater image enhancement based on light-particles interaction descattering
Miao Yang, Arcot Sowmya, Zhiqiang Wei, Bing Zheng · OCEANS 2017 - Aberdeen · 2017
Field optical measurement results show that a classification study of inshore water is necessary. Typical dehazing methods rely on a dark channel model using global rough estimates of back-scattering, that performs poorly on inshore underwater image restoration. This paper introduces a restoration approach to the problem that focuses on inshore water body optical properties. The method introduces transmission map estimation by applying lighting-reflection decomposition rather than replacing the back-scattering with the theoretical maximum. A statistical prior of inshore water body attenuation coefficients is also included in the transmission map estimation. Besides, we propose a novel backscatter lighting estimation method based on the phenomenon of interaction of light rays with particles in the medium. A detailed qualitative analysis against existing techniques is also performed which shows that the proposed method outperforms the state-of-art on inshore underwater data, especially on sharpness and total colour metrics.