Improved Generative Adversarial Network Underwater Image Enhancement Algorithm Based on Multi-channel Features
To address issues such as color deviation,blurry details,and contrast in underwater images,a multi-channel feature enhancement generative adversarial network is proposed for un-derwater image enhancement algorithms.Firstly,to achieve color balance in underwater images,a red channel compensation algorithm is designed by constructing an attenuation matrix between the blue-green channel and the red channel.Secondly,a generative network and a discriminative network is built separately.To obtain feature information at different levels,the generative net-work is composed of feature extraction networks of different depths,and a feature enhancement module is built to enhance multiple features in parallel and improve the quality of the features.At the same time,in order to improve the network's ability to enhance details,a multi-scale discriminative network is built to discriminate the generated images.Finally,multiple loss functions are constructed to optimize network parameters and improve network robustness.The experimental results show that the enhanced image has richer colors and higher contrast compared to the original image.Compared with ex-isting algorithms,this method can effectively restore the true color and texture details of underwater im-ages,significantly improving the quality of underwater images.
NETL
NSTL
万方数据
