Edge-guided feature fusion network for RGB-T salient object detection

Area of Science:

• Computer Vision
• Artificial Intelligence
• Image Processing

Background:

• Salient Object Detection (SOD) in RGB-T images aims to segment important regions across visible and thermal spectra.
• Existing methods often fail to fully leverage the complementary information between RGB and thermal modalities.
• Accurate SOD is crucial for various applications, including surveillance, robotics, and autonomous driving.

Purpose of the Study:

• To propose a novel network, the Edge-Guided Feature Fusion Network (EGFF-Net), for enhanced RGB-T Salient Object Detection.
• To effectively integrate complementary features from RGB and thermal images.
• To improve the accuracy and boundary refinement of salient object segmentation.

Main Methods:

• Cross-modal feature extraction to capture united and intersecting information from RGB and thermal images.
• Edge-guided feature fusion module to enhance salient region details using edge information.
• Layer-by-layer decoding structure for multi-level feature integration and salience map generation.

Main Results:

• EGFF-Net achieved superior performance compared to state-of-the-art methods on three benchmark datasets.
• The proposed modules demonstrated effectiveness in improving both detection accuracy and boundary refinement.
• Extensive experiments validated the robustness and efficacy of the EGFF-Net framework.

Conclusions:

• Integrating cross-modal information and edge-guided fusion is vital for advancing RGB-T SOD.
• EGFF-Net provides a robust framework for multi-modal saliency detection, outperforming existing techniques.
• The findings pave the way for future research in more accurate and refined multi-modal object detection.