An Underwater Image Enhancement Method Based on Multi-branch Generative Adversarial Network

Abstract

The invention discloses an underwater image enhancement method based on a multi-branch generation confrontation network. In the present invention, the underwater degraded original image, the fused underwater clear image under the same scene, and the underwater clear generated image under the same scene are used as training sample sets, and input to the attribute branch network and the discrimination branch network to obtain the attribute map and discrimination picture. The weights of the GAN network are updated by the gradient descent of the respective cost functions of the attribute map and the discriminant map. Until the end of this network training, a model for enhancing degraded underwater images is obtained. The key of the present invention is to imitate the enhancement strategy for underwater images degraded by different factors by using the data-driven characteristics of generative adversarial network and strong imitation ability. A model can be used to solve a variety of underwater image degradation problems caused by different reasons, and the model is more general. The multi-branch structure of attribute branch and discriminative branch is used to enhance the comprehensiveness and robustness of learning.

Description

technical field [0001] The invention belongs to the field of deep learning and relates to an underwater image enhancement method based on a multi-branch generation confrontation network. Background technique [0002] The complex underwater imaging environment and lighting conditions lead to quality degradation of underwater images. Traditional underwater image enhancement and restoration methods are problematic. The method based on the physical model has great limitations in the assumptions and prior knowledge it relies on. It is not adaptable to complex underwater environments, the designed underwater imaging mathematical model is inaccurate, and the model parameter estimation algorithm is complicated. The method of non-physical model ignores the optical properties of underwater imaging, which is easy to introduce color deviation, and the enhanced image is easy to produce oversaturated or undersaturated areas. Existing methods often have a good processing effect on a sing...

AI Technical Summary

Problems solved by technology

The method based on the physical model has great limitations in the assumptions and prior knowledge it relies on. It is not adaptable to complex underwater environments, the designed underwater imaging mathematical model is inaccurate, and the model parameter estimation algorithm is complex.

The method of non-physical model ignores the optical properties of underwater imaging, which is easy to introduce color deviation, and the enhanced image is easy to produce oversaturated or undersaturated areas

Existing methods often have a good processing effect on a single degradation phenomenon, but the overall comprehensiveness and robustness are not ideal, and there are great limitations in practical application

Images