MBFFNet:Multi-branch feature fusion network for underwater image restoration

Abstract

Underwater image is one of the important way for human to obtain oceans information, but the special physical and chemical properties of underwater environment seriously affect the quality of underwater image. To tackle this issue, we propose a multi-branch named MBFFNet that can recover and enhance underwater images simultaneously. We confirm MBFFNet can well recover the clear underwater images from large particle impurities and ocean light spots through experiments on UIEBD datasets. By considering spatial information, it can better complete the prediction of image restoration. It can reduce the amount of network calculations and parameters while maintaining the effect. Feature Fusion Module is proposed to integrate high-level features with low-level features, so that the network can extract enough feature information and better supplement the details to complete the underwater image restoration. In addition, the network can also improve the color performance while maintain the style and spatial texture of the contents. Comparing with other state-of-the-art image recovery and enhancement methods, MBFFNet can obtain optimal or suboptimal quality of the recovered underwater images with less number of parameters, which is valuable in real world applications.