Underwater image enhancement with image colorfulness measure

Due to the absorption and scattering effects of water, underwater images tend suffer from many severe problems, such as low contrast, grayed out colors blurring content. To improve visual quality images, we proposed a novel enhancement model, which is trainable end-to-end neural model. Two parts constitute overall The first one non-parameter layer for preliminary color correction, then second part consisted parametric layers self-adaptive refinement, namely channel-wise linear shift. For better details, contrast colorfulness, this network jointly optimized by pixel-level characteristic-level training criteria. Through extensive experiments on natural scenes, show that method can get high results.