|
Title:
|
FACE RETRIEVAL USING SALIENT FACIAL POINTS |
|
Author(s):
|
Luigi Cinque, Enver Sangineto |
|
ISBN:
|
978-972-8939-30-4 |
|
Editors:
|
Hans Weghorn, Pedro Isaías and Radu Vasiu |
|
Year:
|
2010 |
|
Edition:
|
Single |
|
Keywords:
|
Face Recognition, Facial Landmark Localization, Biometry. |
|
Type:
|
Full Paper |
|
First Page:
|
199 |
|
Last Page:
|
204 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
We present in this article a system for face retrieval based on the relative positions of a set of prefixed anatomically salient facial features (landmarks). Landmarks are extracted from the face image automatically using a two stage approach. In the first phase, a local classification of SIFT descriptors provides a set of landmark candidates. In the second phase, a modified version of the Hausdorff distance is proposed to correct possible errors of the local phase using global shape information. Once the landmark points are extracted, a face image is represented by means of the facial feature positions. A nearest-neighbor approach is used in order to retrieve from a database of suspect individuals those faces whose landmark dispositions best aligns with the landmarks of the input query. It is important to note that landmark selection can be done with every kind of face representation: either drawings or real images are dealt with at the same manner. For this reason, our system is able to compare real images with drawings while common face recognition systems cannot. |
|
|
|
|
|
|
