Trend Survey: Neural Network for Underwater Image Enhancement
Author : Zalak Hitesh Karnik and Shubham Deepak Bhagat
Abstract :
Underwater environment brings unique challenges for imaging due to factors such as light attention, color distortion, and backscatter, which degrade image quality and obstruct visual perception. Traditional image enhancement techniques often find it difficult to handle these challenges effectively. Deep learning has a promising approach to tackle underwater image enhancement tasks by utilizing its ability to learn complex patterns and features directly from data. This review paper provides a comprehensive survey and analysis of deep learning techniques for underwater image enhancement. Highlighting the importance of clear and accurate visual data in various underwater image applications, including biology, oceanology, security etc., We will move to the limitations of existing methods and some approaches which have already overcome some of the limitations. Through an in-depth literature review, we evaluate the evolution of deep learning models specially trained for underwater images including convolution neural network (CNN), Generative Adversarial Networks (GANs) and autoencoders. We discuss the results of these models and highlight the recent developments such as state-of-art methods, transformation and transfer learning. Moreover, we identify key research directions and emerging trends in the field including development of accurate deep learning models for real time underwater image and its enhancement, integration of multimodal data sources for enhanced performance which will focus on future work on the underwater image enhancement. The paper focuses on the latest innovations and challenges in deep learning based underwater image enhancement. It will provide the foundation for many more models. As this path is continuously getting better, there is an opportunity to innovate multiple algorithmic approaches balancing hardware requirements.
Keywords :
Deep Learning, under water images, CNN, Attention, GAN.
