Face Verification across Age Progression

Problem Statement

Given two age separated face images,

  • How accurate the system cn verify whether the two images belong to the same subject (intra-personal) or different subjects (extra-personal).
  • How does age progression affect the verification performance?.

Abstract

Face verification of individuals under the influence of aging effects is a challenging task in the field of computer vision and has gained its attention in recent years. In this work, we study the face verification task by constructing a simple but powerful and effective representation of the face which uses Local Binary Patter (LBP) histograms. The spatial information is incorporated in the representation by constructing a Gaussian pyramid of the face image and computing the LBP histogram at each level of the pyramid which improves the performance of the system. A set of most discriminative features of the face are extracted using the AdaBoost learning algorithm. A strong classifier is built using a set of weak classifiers extracted using the AdaBoost learning algorithm and is used for classification purposes. We performed various experiments on the FGnet aging database and MORPH database to study the effect of aging in face verification task. The results have been compared with other discriminative approaches and shows significant improvement in the performance of the system. The results indicate that there is significant change in performance with smaller age gaps between the images and the performance stabilizes as the age gap gets larger. Also, the facial hari, glasses, etc. provide discriminative cues in face verification.

Flowchart of the Proposed Method

Hierarchical feature extraction

Results

Related Publication(s)