A real-time restoration model method for underwater images based on self-attention mechanism and GAN
An underwater image and attention technology, applied in the field of image processing and deep learning, can solve the problems of color incongruity, limited visible range of underwater images, non-uniform illumination, etc., to achieve the effect of improving visual quality and excellent recovery performance
Active Publication Date: 2022-08-09
GUILIN UNIV OF ELECTRONIC TECH +1
View PDF16 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
The complexity of the marine environment is diverse, and multiple unfavorable factors such as light transmission in water will be seriously attenuated by water absorption, reflection, and scattering. The collected underwater images will inevitably have limited visible range, blurred, Problems such as low contrast, non-uniform lighting, color incongruity, and noise
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0045] A method for real-time restoration of underwater images based on the self-attention mechanism and GAN of the present invention will be described in further detail below in conjunction with specific embodiments.
[0046] The real-time restoration model method of underwater images based on the self-attention mechanism and GAN proposed by the present invention introduces a self-attention module, and the specific implementation steps are as follows:
[0047] S1. Related work:
[0048] S1.1. Generative Adversarial Network Model: GAN is a deep neural network composed of two networks, a generator and a discriminator. Both the generator and the discriminator adopt the idea of zero-sum game in game theory, in which the goal of the generator is to learn the mapping relationship between degraded underwater images and clear underwater images to generate high-quality clear underwater images; The goal is to make the network learn to distinguish generated fake images from real refe...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention relates to a real-time restoration model of underwater images based on a self-attention mechanism and GAN, and belongs to the technical field of deep learning. It takes the generative adversarial network as the basic architecture, the generative network adopts an encoder-decoder structure, generates synthetic images at the original resolution through 9 residual blocks and deconvolution operations, and introduces a self-attention module, which can capture richer high-level features to improve the model performance, in order to preserve the image content and remove underwater noise at the same time, the discriminative network adopts a structure of multi-branch discriminators including an adversarial branch and a criticism branch. The invention ideally solves the problem of low time efficiency of underwater image processing by adding a self-attention module and improving the model structure. The well-trained GAN-RS-based method can adapt to various underwater situations and has excellent real-time processing. performance.
Description
technical field [0001] The present invention relates to the technical field of image processing and deep learning, and more specifically, the present invention is oriented to the restoration task of underwater degraded images. Background technique [0002] In recent years, underwater robots have been widely used in the exploration of marine resources, and the enhancement and restoration technology of underwater degraded images has attracted much attention due to its important significance for ocean exploration, development and utilization. The complexity of the marine environment is diverse, and multiple unfavorable factors, such as light transmission in water, will be seriously attenuated by water absorption, reflection and scattering. Low contrast, non-uniform lighting, color inconsistencies, and noise. The underwater image enhancement and restoration technology aims to repair the degraded underwater images, in order to improve the low contrast, color distortion and blurr...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06T5/00G06N3/04G06N3/08
CPCG06T5/002G06N3/084G06N3/045
Inventor 罗笑南刘瑞关善文秦子钦孙钦李冀吕智张敏
Owner GUILIN UNIV OF ELECTRONIC TECH