Underwater image restoration method based on cyclic generative adversarial network

Abstract

The invention relates to an underwater image restoration method based on a cyclic generative adversarial network, and the method builds a cyclic generative adversarial network (CycleGAN-VGG) based onperception loss, the network is an end-to-end network architecture, and a restored image is directly obtained through an inputted distorted underwater image. According to the invention, the cyclic consistency loss is reduced; vGG perception loss is added to enhance the structure of the CycleGAN, the overall structure of the input image is reserved by optimizing the cyclic consistency loss, and thedifference of the image feature space is compared by perception loss, so that the network can better recover the detail information of the underwater image, perform underwater image recovery and increase the definition of the image. According to the method, the framework of the Wasserstein GAN is adopted, so that the robustness of the method is improved. Meanwhile, in the training and testing stages, the method does not need paired distortion and real image samples, and does not need any underwater image imaging model parameters.

Description

technical field [0001] The invention relates to an underwater image restoration method based on a cyclic generation confrontation network, and belongs to the technical field of image processing. Background technique [0002] In recent years, ocean engineering and research have increasingly relied on underwater imagery taken by autonomous and remotely operated underwater vehicles. However, due to wavelength-dependent light absorption and scattering and the influence of low-end optical imaging equipment, underwater images often suffer from degradations such as insufficient contrast, dispersion, and noise. In addition, the magnitude of light absorption and scattering depends on several complex factors, including water temperature and salinity, and the type and amount of particulate matter in the water. Severe degradation makes it difficult to restore the appearance and color of underwater images. However, color is very important for the study of underwater vision tasks. Ther...

AI Technical Summary

Problems solved by technology

However, due to the complex physical and optical factors underwater, these traditional methods are difficult to realize

At the same time, due to the lack of rich training data, these methods have poor generalization in different underwater environments, and the enhanced images of some scenes often have a tendency to over-enhance or under-enhance

Images