Related Experiment Video
Updated: Jun 11, 2025
10:56
Long-term Behavioral Tracking of Freely Swimming Weakly Electric Fish
Published on: March 6, 2014
12.5K
UICE-MIRNet guided image enhancement for underwater object detection.
Pratima Sarkar1,2, Sourav De3, Sandeep Gurung4
1Department of Computer Science and Engineering, Sikkim Manipal Institute of Technology, Sikkim Manipal University, Rangpo, Sikkim, 737136, India. psmoon2@gmail.com.
Scientific Reports
|September 28, 2024
Summary
This study introduces UICE-MIRNet, an enhanced underwater image technique, to improve object detection in low-light conditions. The method boosts visibility and colorfulness, outperforming existing approaches for marine resource monitoring.
Area of Science:
- Computer Vision
- Marine Biology
- Image Processing
Background:
- Underwater object detection is vital for marine ecosystem monitoring.
- Low-light and scattered lighting conditions pose significant challenges for computer vision systems.
- Existing methods struggle with enhancing visibility and colorfulness in underwater imagery.
Purpose of the Study:
- To propose an enhanced underwater image technique, UICE-MIRNet, for improved object detection.
- To address visibility issues caused by low-light and low-colorfulness in underwater environments.
- To enhance the detection of small, multiple, and dense objects in marine settings.
Main Methods:
- Developed UICE-MIRNet, a specialized version of MIRNet, focusing on brightness and colorfulness enhancement.
- Introduced the Underwater Image-Colorfulness Enhancement Block (UI-CEB) for targeted color correction.
- Integrated UICE-MIRNet with the YOLOv4 object detection model for enhanced performance.
- Utilized convolutional streams, feature fusion, and feature selection for robust image processing.
Main Results:
- UICE-MIRNet effectively enhances colorfulness and restricts brightness in underwater images.
- The enhanced images significantly improved the performance of the YOLOv4 object detection model.
- Quantitative evaluations using UIQM, UCIQE, entropy, and PSNR demonstrated superior performance.
- The proposed method outperformed existing image enhancement and restoration techniques on Brackish and Trash-ICRA19 datasets.
Conclusions:
- UICE-MIRNet offers a robust solution for underwater image enhancement, crucial for accurate object detection.
- The technique successfully addresses challenges posed by poor lighting and low color saturation.
- This advancement supports more effective monitoring of aquaculture resources and marine ecosystems.
