MATLAB comparisons of five PCA methods for face classification and reconstruction with nearest-neighbor classification.
Copyright (C) 2013 Jing Wang
Comparison of five dimensionality reduction algorithms in face classification and reconstruction. The five algorithms are: PCA, PCA-L1, 2DPCA, 2DPCA-L1, 2DPCAL1-S. Classifier is chosen to be Nearest Neighbor(NN).
- demo.m: workflow.
- demo_load_data.m: Yale image loading.
- demo_classification.m and demo_reconstruction.m: evaluation stages.
- yalefaces.zip and data.zip: supplied archives.
Haixian Wang and Jing Wang, "2DPCA with L1-norm for simultaneously robust and sparse modelling," Neural Networks, vol. 46, no. 0, pp. 190-198, 2013.
Run from the repository directory in MATLAB with Image Processing Toolbox available: the loader calls imresize and montage. demo.m extracts the supplied yalefaces.zip, loads 15 subjects with 11 images each, performs classification, adds noise, and runs reconstruction. The loader writes Yale.mat.
Run demo.m.
See the existing GPL-3.0 license.
Jing Wang
wangjing0@seu.edu.cn
yuzhounh@163.com
2013-6-15 20:17:43