|
![](images/pix01.gif) |
|
|
|
Proposing a Features Extraction based on Classifier Selection to Face Recognition and Image Processing |
|
PP: 191-198 |
|
Author(s) |
|
Sajad Parvin,
Zahra Rezaei,
|
|
Abstract |
|
feature is a Gabor response of image with a different tuple (x, k, ). We use a pre-processing whereby we can use a fixed point x for all images without missing of the generality. Eight orientation frequency values are selected for parameter. Five spatial frequency values are also selected for domain of k parameter. So we reach a k Gabor-wavelet based feature space. Also to get rid of the curse of dimensionality problem again without loss of the generality we omit the versatility of values in the k parameter. Indeed we compute the similarity of a pair faces in two images by averaging their similarity defined for all possible values of k parameter for a given parameter. Then considering the similarities of as a matrix we produce eight matrices for eight different parameters. By considering each of these matrices as a classifier we finally use an ensmble mechanism to aggregate them into final classification. We turn to a weighted majority average voting classifier ensemble to handle the problem. We show that the proposed mechanism works well in an employees repository of our laboratory. |
|
|
![](images/pix01.gif) |
|
|