Age Transformation for Improving Face Recognition Performance
This paper presents a novel age transformation algorithm to handle the challenge of facial aging in face recognition. The proposed algorithm registers the gallery and probe face images in polar coordinate domain and minimizes the variations in facial feat
- PDF / 1,797,749 Bytes
- 8 Pages / 430 x 660 pts Page_size
- 15 Downloads / 220 Views
West Virginia Univeristy, Morgantown WV 26506, USA [email protected], [email protected], [email protected] 2 Purvanchal University, Uttar Pradesh 222001, India [email protected]
Abstract. This paper presents a novel age transformation algorithm to handle the challenge of facial aging in face recognition. The proposed algorithm registers the gallery and probe face images in polar coordinate domain and minimizes the variations in facial features caused due to aging. The efficacy of the proposed age transformation algorithm is validated using 2D log polar Gabor based face recognition algorithm on a face database that comprises of face images with large age progression. Experimental results show that the proposed algorithm significantly improves the verification and identification performance.
1
Introduction
Human face undergoes significant changes as a person grows older. The facial features vary for every person and are affected by several factors such as exposure to sunlight, inherent genetics, and nutrition. The performance of face recognition systems cannot contend with the dynamics of temporal metamorphosis over a period of time. Law enforcement agencies such as crime and record bureau regularly require matching a probe image with the individuals in the missing person database. In such applications, there may be significant differences between facial features of probe and gallery images due to age variation. For example, if the age of a probe image is 15 years and the gallery image of the same person is of 5 years, existing face recognition algorithms are ineffective and may not yield the desired results. One approach to handle this challenge is to regularly update the database with recent images or templates. However, this method is not feasible for applications such as border control, homeland security, and missing person identification. To address this issue, researchers have proposed several age simulation and modeling techniques. These techniques model the facial growth over a period of time to minimize the difference between probe and gallery images. Burt and Perrett et al. [1] proposed an age simulation algorithm using shape and texture, and created composite face images for different age groups. They further analyzed and measured the facial cues affected by age variations. Tiddeman [2] proposed wavelet transform based age simulation to prototype the composite face images. Lanitis et al. [3] - [5] proposed statistical models for face simulation. They used A. Ghosh, R.K. De, and S.K. Pal (Eds.): PReMI 2007, LNCS 4815, pp. 576–583, 2007. c Springer-Verlag Berlin Heidelberg 2007
Age Transformation for Improving Face Recognition Performance
577
training images to learn the relationship between coded face representation and actual age of subjects. This relationship was then used to estimate the age of an individual and to reconstruct the face at any age. Gandhi [6] proposed Support Vector Regression to predict the age of frontal faces. He further used the aging function with the image based surface detail tran
Data Loading...
