DCT-based unique faces for face recognition using Mahalanobis distance

Authors:
Vikas Maheshkar;Sushila Kamble;Suneeta Agarwal;Vinay kumar Srivastava
Affiliations:
MNNIT, Allahabad, India;MNNIT, Allahabad, India;MNNIT, Allahabad, India;MNNIT, Allahabad, India
Venue:
Proceedings of the First International Conference on Intelligent Interactive Technologies and Multimedia
Year:
2010

Citing 8
Cited 0

Digital Image Processing

Digital Image Processing
Face Recognition: Features Versus Templates

IEEE Transactions on Pattern Analysis and Machine Intelligence
Face recognition using DCT-based feature vectors

ICASSP '96 Proceedings of the Acoustics, Speech, and Signal Processing, 1996. on Conference Proceedings., 1996 IEEE International Conference - Volume 04
Eigenfaces for recognition

Journal of Cognitive Neuroscience
Learning a Mahalanobis distance metric for data clustering and classification

Pattern Recognition
An image preprocessing algorithm for illumination invariant face recognition

AVBPA'03 Proceedings of the 4th international conference on Audio- and video-based biometric person authentication
A comparison of photometric normalisation algorithms for face verification

FGR' 04 Proceedings of the Sixth IEEE international conference on Automatic face and gesture recognition
Illumination compensation and normalization for robust face recognition using discrete cosine transform in logarithm domain

IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper, we propose a technique to generate DCT based unique normalized face using Principal Component Analysis (PCA). The idea of the PCA is to decompose face images into a small set of characteristic feature images. In the proposed technique we generate feature image by finding the peak values in the absolute DCT matrix followed by normalization. This maximizes the scatter between training dataset to give more discriminating power. The feature images so generated are called unique normalized faces as each image is different and unique from all other training faces. They have high recognition performance since they capture the global features onto a low dimensional linear "face space" extracted from the individual face of training dataset. We use Mahalanobis distance to measure the recognition between original face and the test face. The algorithm is tested on ORL face datasets. In the proposed technique we improved face recognition rate as compared to Eigenface, DCT-normalization and Wavelet-Denoising.