PCA applied to data identifies the directions in the feature space (principal
components) that account for the most variance in the data.
Linear Discriminant Analysis (LDA) tries to identify characteristics that
account for the most variance between classes.
project faces onto a lower dimensional sub-space
no distinction between inter- and intra-class variabilities.
optimal for representation but not for discrimination.
find a sub-space which maximizes the ratio of inter-class and intra-class
same intra-class variability for all classes
There has been a tendency in the computer vision
community to prefer LDA over PCA.
This is mainly because LDA deals directly with
discrimination between classes while PCA does not
pay attention to the underlying class structure.
LDA has lower error rates
LDA works well even if different illumination
LDA works well even if different facial express.
Is LDA always better than PCA?
This paper shows that when the training set is
small, PCA can outperform LDA.
A. Martinez, A. Kak, "PCA versus LDA", IEEE Transactions on Pattern Analysis and
Machine Intelligence, vol. 23, no. 2, pp. 228-233, 2001
Our project also show the same result.
When the number of samples is large and
representative for each class, LDA outperforms
There are five classes (person)
Each class has Eight elements (face images)
1 to 8 in class 1, 9 to 16 in class 2,and so on…
10 test images
Result : PCA and show different face images even using the same training
Efficiency=(Number of correct recognition)/(Total number of test faces)