Underwater image enhancement method based on double U-nets

Abstract

The invention belongs to the field of computer vision, and relates to an underwater image enhancement method based on double U-nets. According to the invention, the method comprises the steps: settingtwo U-net connection convolutional network frameworks based on an underwater imaging physical model, wherein each U-net is composed of an encoding-decoding network, and an attention mechanism is added in front of the first U-net; by processing an input image and a grayscale image thereof, outputting a piece of feature information at each layer of a decoding structure for the second U-net, and estimating a transmission image; processing the underwater image through the second U-net to estimate the underwater image after red light compensation, and connecting the decoding structure with the feature information output by the first U-net to ensure that the details of the image enhancement process are not lost; and finally, dividing the imaging model by the transmission graph to obtain a finalresult. According to the method, the underwater image with serious color distortion and serious atomization can be effectively enhanced, and the details of the image are reserved.

Description

technical field [0001] The invention belongs to the field of computer vision, and relates to an underwater image enhancement method based on a physical model-based dual U-net network. Background technique [0002] With the reduction of terrestrial resources, it is the general trend for human beings to explore the ocean, and underwater images are one of the most intuitive means to reflect the internal conditions of the ocean. In many fields that use underwater images, we need to obtain high-quality underwater images, such as underwater archaeology, underwater biological research, seabed exploration, underwater unmanned submersible vehicle navigation, underwater target detection and tracking and other fields. [0003] The algorithm of underwater image processing technology can be divided into image enhancement algorithm and image restoration algorithm according to whether it is based on the underwater imaging model. The image enhancement method is to study and enhance the pix...

AI Technical Summary

Problems solved by technology

Generally, methods based on physical models that require artificial measurement of prior information have problems such as narrow application range and measurement errors, and the direct mapping of general deep learning methods will lead to problems such as loss of image details. The present invention combines these two methods, both Can improve their respective defects and get better results

Images