Mehmet Koç

A new solution to one sample problem in face recognition using FLDA

Fisher linear discriminant analysis (FLDA) is a very popular method in face recognition. But FLDA fails when one image per person is available. This is due to the fact that the within-class scatter matrices cannot be calculated. An image decomposition method that uses QR-decomposition with column pivoting (QRCP) is proposed in this paper to overcome one image per person problem. At first, the image and its two approximations that are evaluated using QRCP-decomposition are all placed in the training set. Then 2D-FLDA method becomes applicable with these new data. The performance of the proposed image decomposition algorithm is tested on five different face databases, namely ORL, FERET, YALE, UMIST, and PolyU-NIR using 2D-FLDA. Our image decomposition algorithm performs better than the SVD based method mentioned by Gao et al. (2008) [1] in terms of recognition rate and training time in all of the above databases.

Application of the Discriminative Common Vector Approach to one sample problem

Matrix-based (2D) methods have advantages over vector-based (1D) methods. Matrix-based methods generally have less computational costs and higher recognition performances with respect to vector-based variants. In this work a two dimensional variation of Discriminative Common Vector Approach (2D-DCVA) is implemented. The performance of the method in single image problem is compared with the one dimensional Discriminative Common Vector Approach (1D-DCVA) and the two dimensional Fisher Linear Discriminant Analysis (2D-FLDA) on ORL, FERET, and YALE face databases. The best recognition performances are achieved in all databases with the proposed method.

Feature selection method with common vector and discriminative common vector approaches

The dimension of the feature vector is very important for real time face recognition applications. High dimensional feature vectors increase the computational complexity and execution time of the face recognition system. In this work, a new feature selection method is proposed related with CVA and DCVA to reduce the dimension of the face images. Experiments are executed on two different face databases, namely AR, FERET. Great dimension reduction is achieved with slight recognition rate loss.

Some variants of spiral LBP in texture recognition

Texture classification is one of the recently popular study topics in pattern recognition. Local binary pattern (LBP) is a very efficient local texture descriptor and is used for feature extraction in texture recognition. There are five main steps in representation of texture images: neighbourhood topology and sampling, thresholding and quantisation, encoding and regrouping, combining complementary features. In this study, the authors used symmetric two spirals LBP to measure the grey-scale difference between the centre pixel and its neighbours. They also extended the proposed method by using four spirals LBP to generate the LBP code. For classification, linear regression classification method, which is generally used to solve the face recognition problems, is used. The authors tested the performance of their method on UIUC and CUReT texture image databases. It is experimentally demonstrated that the proposed method achieves the highest classification accuracy among the comparative methods on texture databases.

A comparison of classification methods for local binary patterns

Texture recognition is an important tool used for content-based image retrieval, face recognition, and satellite image classification applications. One of the most successful features for texture recognition is local binary patterns (LBP), which computes local intensity differences for a pixel with respect to its neighbor pixels. In many studies in the literature, histogram based similarity measures are employed to classify LBP features. In this study, we investigate the performance of support vector machines, linear discriminant analysis, and linear regression classifier to improve the success of LBP features. We achieved 84.4% classification success using linear regression classification.

A NOVEL IMPLEMENTATION ALGORITHM FOR CALCULATION OF COMMON VECTORS

Common vector approach (CVA), discriminative common vector approach (DCVA) and linear regression classification (LRC) are subspace methods used in pattern recognition. Up to now, there were two well-known algorithms to calculate the common vectors: ( i ) by using the Gram-Schmidt orthogonalization process, ( ii ) by using the within-class covariance matrices. The purpose of this paper is to introduce a new implementation algorithm for the derivation of the common vectors using the linear regression idea. The derivation of the discriminative common vectors through LRC is also included in this paper. Two numerical examples are given to clarify the proposed derivations. An experimental work is given in AR face database to compare the recognition performances of CVA, DCVA, and LRC. Additionally, the three implementation algorithms of common vector are compared in terms of processing time efficiency.

Modular common vector approach

The performance of a face recognition system is negatively affected by the accessories used on the face. Like many methods, the recognition performance of the Common Vector Approach (CVA) [1] over occluded images is not at the desired level. In this work, we proposed an extension of the CVA, namely the Modular Common Vector Approach (M-CVA), which improves the recognition performance at the occluded face images. M-CVA outperforms CVA by a margin of 82, 7 percent in the experiments which are conducted over AR face database.

A fast method for the Common Vector Approach

In this paper a new method is proposed to perform the Common Vector Approach(CVA). While CVA performs the classification with respect to the distance between vectors, the new method performs the classification with respect to the scalars. In experimental work, AR face database is used. Method performs the classification approximately 2 times faster than the classical calculations in AR face database.