Towards underwater image enhancement using super-resolution convolutional neural networks

Abstract

Underwater objects detection and recognition is challenging because of the degradation of underwater images, such as color casts, blurring and low contrast. To tackle this problem, a novel underwater image enhancement method is proposed. It consists of two main steps. First, an adaptive color correction algorithm is used to compensate color casts and produce natural color corrected images. Second, a super-resolution convolutional neural network is applied to color corrected images in order to remove blurring. The proposed network learns a relationship which can be employed into image de-blurring from a large amount of blurry images and the corresponding clear images. Based on the relationship, the color corrected image will be de-blurred and sharpened. The experimental results show that the proposed strategy improves the quality of underwater images efficiently and arrives at good results in underwater objects detection and recognition.