In this paper, we present a method for face recognition using boosted Gabor feature based classifiers. Weak classifiers are constructed based on both magnitude and phase features derived from Gabor filters [Quadrature-phase simple-cell pairs are ap-propriately described in complex analytic from]. The multi-class problem is transformed into a two-class one of intra- and extra-class classification using intra-personal and extra-personal difference images, as in [Beyond euclidean eigenspaces:bayesian matching for visian recognition]. A cascade of strong classifiers are learned using bootstrapped negative examples, similar to the way in face detection framework [Robust real time object detection]. The combination of classifiers based on two different types of features produces better results than using either type. Experiments on FERET database show good results comparable to the best one reported in literature [The FERET evaluation methodology for face-recognition algorithms].
Modeling subspaces of a distribution of interest in high dimensional spaces is a challenging problem in pattern analysis. In this paper, we present a novel framework for pose invariant face detection through. multi-view face distribution modeling. The approach is aimed to learn a set of low-dimensional subspaces from an originally nonlinear distribution by using the mixtures of probabilistic PCA. From the experiments, we found the learned PPCA models are of low dimensionality and exhibit high local linearity, and consequently offer an efficient representation for visual recognition. The model is then used to extract features and select "representative" negative training samples. Multi-view face detection is performed in the derived feature space by classifying each face into one of the view classes or into the nonface class, by using a multi-class SVM array classifier. The classification results from each view are fused together and yields the final classification results. The experimental results demonstrate the performance superiority of our proposed framework while performing multi-view face detection.
Active appearance models (AAM) is very powerful for extracting objects, e.g. faces, from images. It is composed of two parts: the AAM subspace model and the AAM search. While these two parts are closely correlated, existing efforts treated them separately and had not considered how to optimize them overall. In this paper, an approach is proposed to optimize the subspace model while considering the search procedure. We first perform a subspace error analysis, and then to minimize the AAM error we propose an approach which optimizes the subspace model according to the search procedure. For the subspace error analysis, we decomposed the subspace error into two parts, which are introduced by the subspace model and the search procedure respectively. This decomposition shows that the optimal results of AAM can be achieved only by optimizing both of them jointly rather than separately. Furthermore, based on this error decomposition, we develop a method to find the optimal subspace model according to the search procedure by considering both the two decomposed errors. Experimental results demonstrate that our method can find the optimal AAM subspace model rapidly and improve the performance of AAM significantly.
This paper presents an approach to face alignment under variable illumination, an obstacle largely ignored in previous 2D alignment work. To account for illumination variation, our method employs two forms of relatively lighting-invariant information. Edge phase congruency is adopted to coarsely localize facial features, since prominent feature edges can be robustly located in the presence of shading and shadows. To accurately deal with features with less pronounced edges, final alignment is then computed from intrinsic gray-level information recovered using a proposed form of local intensity normalization. With this approach, our face alignment system works efficiently and effectively under a wide range of illumination conditions, as evidenced by extensive experimentation.
In this paper, we propose a novel method, called local non-negative matrix factorization (LNMF), for learning spatially localized, parts-based subspace representation of visual patterns. An objective function is defined to impose a localization constraint, in addition to the non-negativity constraint in the standard NMF. This gives a set of bases which not only allows a non-subtractive (part-based) representation of images but also manifests localized features. An algorithm is presented for the learning of such basic components. Experimental results are presented to compare LNMF with the NMF and PCA methods for face representation and recognition, which demonstrates advantages of LNMF.
Independent subspace analysis (ISA) is able to learn view-subspaces unsupervisedly from (view-unlabeled) multi-view face examples (S.Z. Li et al., 2001). We explain underlying reasons for the emergent formation of ISA view-subspaces. Based on the analysis, we present a supervised method for more effective learning of view-subspace, assuming that view-labeled face examples are available. The models thus learned give more accurate pose estimation than those obtained with the unsupervised ISA.
Face representation based on Gabor features has attracted much attention and achieved great success in face recognition area for the advantages of the Gabor features. However, Gabor features currently adopted by most systems are redundant and too high dimensional. In this paper, we propose a face recognition method using AdaBoosted Gabor features, which are not only low dimensional but also discriminant. The main contribution of the paper lies in two points: (1) AdaBoost is successfully applied to face recognition by introducing the intra-face and extra-face difference space in the Gabor feature space; (2) an appropriate re-sampling scheme is adopted to deal with the imbalance between the amount of the positive samples and that of the negative samples. By using the proposed method, only hundreds of Gabor features are selected. Experiments on FERET database have shown that these hundreds of Gabor features are enough to achieve good performance comparable to that of methods using the complete set of Gabor features.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)