Related Experiment Video
Updated: Aug 17, 2025
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
1.9K
GMLM-CNN: A Hybrid Solution to SWIR-VIS Face Verification with Limited Imagery
Zhicheng Cao1, Natalia A Schmid2, Shufen Cao3
1Molecular and Neuroimaging Engineering Research Center of Ministry of Education, School of Life Science and Technology, Xidian University, Xi'an 710071, China.
Sensors (Basel, Switzerland)
|December 11, 2022
Summary
This study introduces GMLM-CNN, a hybrid approach for cross-spectral face verification using short-wave infrared (SWIR) and visible light (VIS) images. It effectively handles limited SWIR data, outperforming existing methods.
Area of Science:
- Computer Science
- Biometrics
- Image Processing
Background:
- Cross-spectral face verification (e.g., short-wave infrared to visible light) is challenging due to differing image characteristics.
- Real-world applications like nighttime surveillance require robust face recognition across spectra, especially with limited data in bands like SWIR.
Purpose of the Study:
- To develop a hybrid method combining traditional feature engineering and deep learning for improved cross-spectral face verification.
- To address the challenge of limited imagery in the short-wave infrared (SWIR) spectrum for face recognition.
Main Methods:
- Introduced two new measurement level operators: Nominal Measurement Descriptor (NMD) and Interval Measurement Descriptor (IMD).
- Proposed a composite operator, Gabor Multiple-Level Measurement (GMLM), to fuse features from multiple measurement levels.
- Developed a GMLM-CNN framework integrating GMLM features with a Principal Component Analysis (PCA)-based neural network for feature selection and recognition.
Main Results:
- The GMLM-CNN framework demonstrated superior performance in cross-spectral face verification compared to traditional hand-crafted operators and state-of-the-art deep learning models.
- Experimental results on a dataset of VIS and SWIR faces confirmed the effectiveness of the hybrid approach, particularly under limited data conditions.
Conclusions:
- The proposed GMLM-CNN hybrid method offers a significant advancement in cross-spectral face verification, especially when dealing with limited SWIR imagery.
- This approach provides a robust solution for applications requiring face recognition across different spectral bands, enhancing surveillance and security capabilities.
