Proceedings of the
The Nineteenth International Conference on Computational Intelligence and Security (CIS 2023)
December 1 – 4, 2023, Haikou, China
An Improved Shallow-UWnet Based Image Enhancement Method for Turbid Water
School of Information and Control Engineering, Qingdao University of Technology, China.
ABSTRACT
In turbid water environments, the phenomena of reduced underwater image contrast, color deviation, and blurriness are caused by light absorption and scattering. To address the significant degradation in image quality of underwater images in turbid conditions, we proposed a method based on an improved Shallow-UWnet network model for enhancing underwater images. To adapt to the characteristics of imaging in turbid water, an underwater turbid image dataset is constructed through experimental simulations of real turbid water conditions. Firstly, the original images are globally color-corrected using the gray-world algorithm, and then the processed underwater turbid image dataset is fed into the improved Shallow-UWnet network model for learning, resulting in enhanced images. Experimental results indicate that this method substantially improves image quality and achieves underwater image enhancement in various turbid water scenarios. Furthermore, significant progress is achieved in both subjective and objective evaluation metrics. This method holds great potential for widespread applications in domains such as underwater rescue and underwater exploration.
Keywords: Image processing, Underwater image enhancement, Turbid water, Deep learning, Improved Shallow-UWnet network model.
Download PDF
School of Information and Control Engineering, Qingdao University of Technology, China.
ABSTRACT
In turbid water environments, the phenomena of reduced underwater image contrast, color deviation, and blurriness are caused by light absorption and scattering. To address the significant degradation in image quality of underwater images in turbid conditions, we proposed a method based on an improved Shallow-UWnet network model for enhancing underwater images. To adapt to the characteristics of imaging in turbid water, an underwater turbid image dataset is constructed through experimental simulations of real turbid water conditions. Firstly, the original images are globally color-corrected using the gray-world algorithm, and then the processed underwater turbid image dataset is fed into the improved Shallow-UWnet network model for learning, resulting in enhanced images. Experimental results indicate that this method substantially improves image quality and achieves underwater image enhancement in various turbid water scenarios. Furthermore, significant progress is achieved in both subjective and objective evaluation metrics. This method holds great potential for widespread applications in domains such as underwater rescue and underwater exploration.
Keywords: Image processing, Underwater image enhancement, Turbid water, Deep learning, Improved Shallow-UWnet network model.
Download PDF