A Probabilistic Fusion Methodology for Face Recognition
- PDF / 2,863,062 Bytes
- 16 Pages / 600 x 792 pts Page_size
- 49 Downloads / 252 Views
A Probabilistic Fusion Methodology for Face Recognition K. Srinivasa Rao Image Processing and Computer Vision Laboratory, Department of Electrical Engineering, Indian Institute of Technology Madras, Chennai-600 036, India Email: srinu [email protected]
A. N. Rajagopalan Image Processing and Computer Vision Laboratory, Department of Electrical Engineering, Indian Institute of Technology Madras, Chennai-600 036, India Email: [email protected] Received 11 May 2004; Revised 30 March 2005; Recommended for Publication by Satya Dharanipragada We propose a novel probabilistic framework that combines information acquired from different facial features for robust face recognition. The features used are the entire face, the edginess image of the face, and the eyes. In the training stage, individual feature spaces are constructed using principal component analysis (PCA) and Fisher’s linear discriminant (FLD). By using the distance-in-feature-space (DIFS) values of the training images, the distributions of the DIFS values in each feature space are computed. For a given image, the distributions of the DIFS values yield confidence weights for the three facial features extracted from the image. The final score is computed using a probabilistic fusion criterion and the match with the highest score is used to establish the identity of a person. A new preprocessing scheme for illumination compensation is also advocated. The proposed fusion approach is more reliable than a recognition system which uses only one feature, trained individually. The method is validated on different face datasets, including the FERET database. Keywords and phrases: face recognition, block histogram modification, edginess image, probabilistic fusion, distance in feature space.
1.
INTRODUCTION
Automatic face recognition is a challenging problem in computer vision. Computers that recognize faces can be applied to a wide variety of problems such as information mining from face databases, security systems, and humancomputer interactions [1]. Existing face recognition methodologies may be broadly classified into two categories, holistic and analytic [2]. Here, we refer to some well-known works in both categories. The survey is by no means exhaustive. Holistic approaches consider global properties of the face. The eigenface-based face recognition system proposed by Turk and Pentland [3] uses principal component analysis (PCA) to compute linear projections of face images to arrive at a compact representation. The method has good recognition rate but is sensitive to variations in facial expressions and ambient illumination. For higher discriminability, Fisher’s linear discriminant (FLD) analysis in conjunction with PCA has been proposed in the literature [4, 5]. Elastic graph matching for face recognition is yet another popular approach [6]. It uses a novel dynamic-link architecture
based on multiscale morphological dilation and erosion for authentication of frontal images. The idea is to weight the graph nodes according to their discriminating power. Duc e
Data Loading...
