Race classification using gaussian-based weight K-nn algorithm for face recognition
One of the greatest challenges in facial recognition systems is to recognize faces around different race and illuminations. Chromaticity is an essential factor in facial recognition and shows the intensity of the color in a pixel, it can greatly vary depending on the lighting conditions. The race c...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
University of Kuwait
2018
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/85404/1/SasanKaramizadeh2018_RaceClassificationUsingGaussianBasedWeight.pdf http://eprints.utm.my/id/eprint/85404/ https://kuwaitjournals.org/jer/index.php/JER/article/view/3062 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Universiti Teknologi Malaysia |
| Language: | English |
| Summary: | One of the greatest challenges in facial recognition systems is to recognize faces around different race and illuminations. Chromaticity is an essential factor in facial recognition and shows the intensity of the color in a pixel, it can greatly vary depending on the lighting conditions.
The race classification scheme proposed which is Gaussian based-weighted K-Nearest Neighbor classifier in this paper, has very sensitive to illumination intensity. The main idea is first to identify the minority class instances in the training data and then generalize them to Gaussian function as concept for the minority class. By using combination of K-NN algorithm with Gaussian formula for race classification. In this paper, image processing is divided into two phases. The first is preprocessing phase. There are three preprocessing comprises of auto contrast balance, noise reduction and auto-color balancing. The second phase is face processing which contains six steps; face detection, illumination normalization, feature extraction, skin segmentation, race classification and face recognition. There are two type of dataset are being used; first FERET dataset where images inside this dataset involve of illumination variations. The second is Caltech dataset which images side this dataset contains noises. |
|---|