An ensemble deep learning approach for underwater image enhancement

Abstract

Underwater images taken with underwater cameras often experience visual deterioration, including problems like color distortion, diminished contrasts, and blurred details. The current trend in research is to address these issues separately, resulting in challenges in achieving consistent improvement in overall underwater image clarity. Additionally, this approach often leads to over-enhancement, over-saturation of specific areas, and uneven coloring in the texture of the images. The main focus of this research work is to introduce an ensemble deep learning approach, spatial approach along with deep learning method for underwater image enhancement aimed at improving the image brightness, sharpening, and reducing over-contrast amplification while maintaining the structure of the image. The generalization of proposed Convolutional Neural Network model is proved with respect to performance across different datasets such as UEIB and EUVP datasets and comparison against several models.

Keywords

• Deep learning
• underwater image enhancement
• poor and good quality underwater images
• CNN models