MutualBoost learning for selecting Gabor features for face recognition

Abstract

This paper describes an improved boosting algorithm, the MutualBoost algorithm, and its application in developing a fast and robust Gabor feature based face recognition system. The algorithm uses mutual information to eliminate redundancy among Gabor features selected using the AdaBoost algorithm. Selected Gabor features are then subjected to Generalized Discriminant Analysis (GDA) for class separability enhancement before being used for face recognition. Compared with one of the top performers in the 2004 face verification competition, our method demonstrates clear advantages in classification accuracy, memory and computation. The method has been tested on the whole FERET database using the FERET evaluation protocol. Significant improvement in performance is observed. For example, existing Gabor based methods use a huge number of Gabor features, our method needs only hundreds of Gabor features to achieve very high classification accuracy. Due to substantially reduced feature dimension, memory and computation costs are reduced significantly -only 4 s are needed to recognize 200 face images.

Face recognition using decimated redundant discrete wavelet transforms

“…For the FERET database, we compare DRDWT with HEC [7], aguar'a06NF [35], LGPDP [6], HGPP [37], Atan K-HLGPP [38], and MutualGabor-GDA [5]. The corresponding experimental results are presented in Table 5.…”

“…Until now, a large number of face recognition methods, such as appearance-based methods [1][2][3][4], Gabor wavelet methods [5][6][7]35,38], and machine learning-based methods [8,9], have been developed for still images. Zhao et al [10] provided a comprehensive survey of the studies of machine recognition of faces, categorized existing recognition techniques, and presented a detailed description of a number of representative methods.…”

Gabor wavelets and General Discriminant Analysis for face identification and verification

“…Though the robustness and accuracy of Gabor + GDA has been extensively tested and evaluated we are currently developing methods to reduce the feature dimension further, using high level feature selection schemes such as boosting [34,35], mutual information [36], etc. …”

Factored principal components analysis, with applications to face recognition

“…A total of 200 individuals are selected from the FERET database (Phillips et al 2000; Shen and Bai 2006) with three frontal photographs of the face per individual with size 128 × 128 pixels. One of the images is selected at random and used as a test image whereas the other two images are used as training images.…”

“…3. We use the Gabor wavelet transformation at five scales and eight angles (as in Shen and Bai 2006) We consider the use of a classification method which combines the results from the different Gabor transforms and the raw photos. We combine measures by multiplying the likelihood ratios for each type of image/transform for FLR, or summing up the SVM objective functions for each type of image/transform using SVM.…”