Related Experiment Video
Updated: May 20, 2026
07:45
Quantifying Intermembrane Distances with Serial Image Dilations
Published on: September 28, 2018
Manifold-manifold distance and its application to face recognition with image sets
Ruiping Wang1, Shiguang Shan, Xilin Chen
1Key Laboratory of Intelligent Information Processing of Chinese Academy of Sciences (CAS), Institute of Computing Technology, CAS, Beijing 100190, China. rpwang@mail.tsinghua.edu.cn
Summary
This study introduces manifold-manifold distance (MMD) for face recognition from image sets. MMD effectively measures similarity between image sets, outperforming other methods.
Area of Science:
- Computer Science
- Machine Learning
- Pattern Recognition
Background:
- Face recognition from image sets presents challenges due to large intra-class variations.
- Existing methods often struggle to capture the complex structure within image sets.
Purpose of the Study:
- To propose a novel framework for classifying image sets in face recognition.
- To introduce manifold-manifold distance (MMD) as a general set similarity measure.
Main Methods:
- Modeling image sets as manifolds and formulating classification as computing manifold-manifold distance (MMD).
- Developing a general multilevel MMD framework to handle point, subspace, and manifold levels of image set representation.
- Representing manifolds as collections of local linear models (subspaces) and integrating pairwise subspace distances.
Main Results:
- The proposed MMD framework effectively measures similarity between image sets.
- MMD consistently outperforms competing non-discriminative methods in face recognition tasks.
- MMD shows promising comparability to state-of-the-art discriminative methods.
Conclusions:
- Manifold-manifold distance (MMD) offers a robust approach for face recognition using image sets.
- The multilevel MMD framework provides a versatile tool for set similarity measurement.
- MMD demonstrates significant potential for improving face recognition system performance.
