Facial age range estimation using geometric ratios and hessian-based filter wrinkle analysis

Human Face holds important amount of information such as expression, identity, gender and age. The vast majority of people are able to easily recognize human traits like emotional states, where they can tell if the person is happy, sad or angry from face. However, estimating person’s age from face image is a challenging task. The facial age estimation method has recently gained attention from the computer vision and computer graphic community due to its applications as well as the challenges in the development process. Traditionally, researchers using numerous of ratios obtained from extracted facial features landmark points to measure facial age. Most of those points are obtained from publicly facial aging database. Although the estimation result promising, the method still have limitation because it’s work with manual calibration to detect, to extract all the landmark point to estimate human facial age. Lately, many researchers combine facial features geometric ratios with facial skin texture to estimates human facial age range. Facial skin texture was obtained based on the lines that form wrinkles in the facial area. Based on literature study, a technique often used for facial wrinkles analysis is obtained based on Canny Edge Detector. However, it produces inconsistence performance because edge detector only detect wrinkle boundaries rather than the wrinkle itself. In this thesis, a new automatic facial age range estimation method using geometric ratios and wrinkle analysis is proposed. The geometric ratios are based on combination of facial features distances and angles distribution between selected face features using minimum extracted facial features landmark points. The Hessian-Based Filter is used to enhance wrinkle analysis for age range estimation method. In addition, this research proposed a new algorithm to measure face region end points which also used as landmark points derived from Ideal Frontal Symmetry and Proportion of the Face to estimation age range. The age range was classified using SVM and Multi-SVM classifier and the performance evaluation was tested on FG-NET database. Experiments for each phase in the research framework were qualitatively and quantitatively evaluated. The overall findings show that the proposed method is significantly increase the estimation rate with 92% of accuracy compared to previous methods. The proposed method also able to estimate age of person with no hair or hair that covers part of the forehead. Besides, this work is also successfully implemented in real-time face tracking application because using fully automatic extraction and localization approach.

Preview Text FSKTM 2016 39 IR.pdf Download (557kB) | Preview

Actions (login required)

View Item