Age invariant face recogntion system using automated voronoi diagram segmentation
One of the challenges in automatic face recognition is to achieve sequential face invariant. This is a challenging task because the human face undergoes many changes as a person grows older. In this study we will be focusing on age invariant features of a human face. The goal of this study is to inv...
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2013
|
| Subjects: | |
| Online Access: | http://umpir.ump.edu.my/id/eprint/13520/ http://umpir.ump.edu.my/id/eprint/13520/ http://umpir.ump.edu.my/id/eprint/13520/1/NIK%20NURUL%20AIN%20NIK%20SUKI.pdf |
| Summary: | One of the challenges in automatic face recognition is to achieve sequential face invariant. This is a challenging task because the human face undergoes many changes as a person grows older. In this study we will be focusing on age invariant features of a human face. The goal of this study is to investigate the face age invariant features that can be used for face matching, secondly is to come out with a prototype of matching scheme that is robust to the changes of facial aging and finally to evaluate the proposed prototype with the other similar prototype. The proposed approach is based on automated image segmentation using Voronoi Diagram (VD) and Delaunay Triangulations (DT). Later from the detected face region, the eyes will be detected using template matching together with DT. The outcomes, which are list of five coordinates, will be used to calculate interest distance in human faces. Later ratios between those distances are formulated. Difference vector will be used in the proposed method in order to perform face recognition steps. Datasets used for this research is selected images from FG-NET Aging Database and BioID Face Database, which is widely being used for image based face aging analysis; consist of 15 sample images taken from 5 different person. The selection is based on the project scopes and difference ages. The result shows that 11 images are successfully recognized. It shows an increase to 73.34% compared to other recent methods. |
|---|