Related Experiment Video
Updated: May 6, 2026
03:31
Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications
Published on: December 15, 2023
1.2K
Feature pyramid serial attention network for underwater image enhancement.
Applied Optics
|March 17, 2026
Summary
Underwater images suffer from poor color and illumination. FPSANet, a novel feature pyramid serial attention network, enhances underwater images by fusing multiscale features and using serial attention, improving vision tasks.
Area of Science:
- Computer Vision
- Image Processing
Background:
- Underwater images present challenges like color distortion and low illumination.
- These issues degrade performance in underwater vision applications such as object detection and segmentation.
Purpose of the Study:
- To introduce FPSANet, a feature pyramid serial attention network for underwater image enhancement.
- To address limitations in underwater image quality for improved vision system performance.
Main Methods:
- Implemented a feature pyramid fusion module for integrating multiscale spatial information.
- Developed a serial attention module (SAM) combining pixel, channel, and space attention to prioritize illumination and color details.
Main Results:
- FPSANet demonstrated superior performance across various underwater datasets.
- Achieved at least 3.23% and 1.46% increases in PSNR and SSIM values, respectively, compared to the second-best method.
Conclusions:
- FPSANet effectively enhances underwater images, overcoming common distortions and illumination issues.
- The method shows significant improvements in tasks including image segmentation, keypoint detection, and foggy image enhancement.
