Underwater image enhancement method based on multi-branch network superposition

CN119963436APending Publication Date: 2025-05-09FUJIAN UNIV OF TECH
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202510016164.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-01-06
Publication Date
2025-05-09

AI Technical Summary

Technical Problem

The existing underwater image enhancement technology has problems such as poor results based on non-physical models, difficult methods based on physical models, and deep learning-based methods require a large amount of data and high computing resources.

Method used

Using the multi-branch network overlay method, by constructing an image data set containing the training set and the test set, using U-Net and convolutional neural network modules to build an initial image enhancement model, design branches with different number of channels, and deploy SE attention mechanism modules and convolutional LSTM units to perform image feature extraction, enhancement and deep feature extraction, and denoising through feature overlay combination.

Benefits of technology

Real color reduction, noise removal, increased visibility, image quality and clarity improvement of underwater images, improved target recognition and navigation capabilities, and was suitable for marine biology research and environmental monitoring.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119963436A_ABST
    Figure CN119963436A_ABST
Patent Text Reader

Abstract

The invention discloses an underwater image enhancement method based on multi-branch network superposition, which belongs to the technical field of digital image processing, and comprises the following steps: S1, constructing an image data set comprising a training set and a test set, S2, constructing an initial image enhancement model, S3, inputting the training set in the image data set into the initial image enhancement model, and S4, constructing an image enhancement model; the method comprises the steps of S1, obtaining an initial image enhancement model, S2, obtaining an underwater image enhancement model, S4, training the initial image enhancement model by using a training set to obtain an underwater image enhancement model, S5, testing the underwater image enhancement model by using a test set in an image data set, and S6, evaluating the underwater image enhancement model. Noise of the underwater image is removed, visibility is improved, and meanwhile the quality and definition of the image are improved.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Cited By

  • Underwater image enhancement method and system based on combination of multiple convolution models and attention mechanism

    CN120278912A