Jointly adversarial networks for wavelength compensation and dehazing of underwater images

Xianping Fu, Xueyan Ding, Zheng Liang, Yafei Wang

Abstract: Severe color casts, low contrast, and blur of underwater images caused by light absorption and scattering result in a challenging task for exploring underwater environments. In this paper, we propose a novel joint wavelength compensation and dehazing network (JWCDN) to remove the degradation effects caused by light absorption and scattering. The joint learning framework is based on an improved underwater image formation model that takes into account the vertical (water surface-object path) and horizontal (object-camera path) effects of light propagation in water. By embedding the underwater image formation model into a generative adversarial network, we can jointly estimate the transmission map, wavelength attenuation, and background light via different network modules, and use the underwater image formation model to recover degraded underwater images. To further improve the recovered image, we use an edge-preserving network module to enhance the detail of the recovered image. Moreover, to train the proposed network, we present an effective underwater image synthesis method that generates underwater images using the optical attenuation coefficients of different water types. The synthesis method can simulate the color, contrast, and blur of different underwater imaging environments. Extensive experiments on synthetic and real-world underwater images demonstrate that the proposed method yields comparable or better results on both subjective and objective assessments, compared with several state-of-the-art methods.

External IDs:dblp:journals/mta/FuDLW23