Proceedings of the
The Nineteenth International Conference on Computational Intelligence and Security (CIS 2023)
December 1 – 4, 2023, Haikou, China
Algorithm for Enhancing Turbid Water Underwater Images Based on Lightweight Neural Network
School of Information and Control Engineering, Qingdao University of Technology, China.
ABSTRACT
This paper proposes a lightweight neural network-based model for enhancing underwater images captured in turbid water, which addresses the low efficiency of existing methods in this scenario and the issues of low image clarity, color distortion, and texture loss after enhancement. The proposed model first extracts features from the underwater images captured in turbid water using a lightweight reflection prediction network, which obtains background reflection information and predicts the reflection component of light. Then, a lightweight direct attenuation index network is introduced, which extracts features of the direct component of light and utilizes an attention mechanism to improve the accuracy of predicting the exponent parameter of the direct component. Finally, the enhanced underwater images are generated using an image correction model. Experimental results show that the proposed method achieves good performance in terms of enhancing image clarity, recovering details, and correcting color, compared with other underwater image enhancement methods. Our method performs at a higher speed compared to the state of the art, while maintaining similar or better image quality indicators such as PSNR and SSIM.
Keywords: Lightweight neural network, Rapid enhancement, Turbid water, Image restoration.
Download PDF
School of Information and Control Engineering, Qingdao University of Technology, China.
ABSTRACT
This paper proposes a lightweight neural network-based model for enhancing underwater images captured in turbid water, which addresses the low efficiency of existing methods in this scenario and the issues of low image clarity, color distortion, and texture loss after enhancement. The proposed model first extracts features from the underwater images captured in turbid water using a lightweight reflection prediction network, which obtains background reflection information and predicts the reflection component of light. Then, a lightweight direct attenuation index network is introduced, which extracts features of the direct component of light and utilizes an attention mechanism to improve the accuracy of predicting the exponent parameter of the direct component. Finally, the enhanced underwater images are generated using an image correction model. Experimental results show that the proposed method achieves good performance in terms of enhancing image clarity, recovering details, and correcting color, compared with other underwater image enhancement methods. Our method performs at a higher speed compared to the state of the art, while maintaining similar or better image quality indicators such as PSNR and SSIM.
Keywords: Lightweight neural network, Rapid enhancement, Turbid water, Image restoration.
Download PDF