Zhou, Feng, Dai, Ju, Xia, Shaoyan, Rosin, Paul L.  ORCID: https://orcid.org/0000-0002-4965-3884 and Lai, Yukun ORCID: https://orcid.org/0000-0002-2094-5680
2025.
A survey of RGB-D-based semantic segmentation.
Chen, C. H., ed.
Pattern Recognition and Computer Vision in the New AI Era,
World Scientific,
pp. 41-67.
(10.1142/9789819807154_0003)
|
Abstract
This chapter focuses on semantic segmentation of RGB-D images. RGB-D-based semantic segmentation is a research area that has gained significant attention in the computer vision community. It involves leveraging both color (RGB) and depth (D) information to accurately segment objects in a scene. RGB-D sensors, such as Microsoft’s Kinect or Intel’s RealSense, provide synchronized color and depth measurements of the scene, which offer complementary information. Color provides high-level visual cues, while depth provides detailed geometric information. By fusing both modalities, RGB-D-based semantic segmentation algorithms can achieve more robust and accurate results than from either modality alone. In this chapter, we give a survey about the task of RGB-D-based semantic segmentation. We first review the background of this task. Next, we summarize a taxonomy of scene RGB-D semantic segmentation methods, including historical background, RGB-based semantic segmentation methods, the three main fusion approaches (namely early fusion, middle fusion, and late fusion), and the current attention-based methods and multi-modal methods. We also present several semantic segmentation methods applied to RGB-D-based medical images, including applications involving retinal vessels, coronary arteries, and multi-modal vascular scenarios, as well as unsupervised learning for segmentation based on reconstruction and unsupervised clustering methods. Additionally, we review relevant datasets, evaluation metrics, and experimental results of these methods. Finally, we discuss the limitations in this field and propose some possible directions for future research.
| Item Type: | Book Section |
|---|---|
| Date Type: | Publication |
| Status: | Published |
| Schools: | Schools > Computer Science & Informatics |
| Publisher: | World Scientific |
| ISBN: | 9789819807147 |
| ISSN: | 2010-2143 |
| Last Modified: | 12 Aug 2025 11:00 |
| URI: | https://orca.cardiff.ac.uk/id/eprint/180388 |
Actions (repository staff only)
 |
Edit Item |
Dimensions
Dimensions