Heather M. Liddell

Support vector regression and classification based multi-view face detection and recognition

A support vector machine-based multi-view face detection and recognition framework is described. Face detection is carried out by constructing several detectors, each of them in charge of one specific view. The symmetrical property of face images is employed to simplify the complexity of the modelling. The estimation of head pose, which is achieved by using the support vector regression technique, provides crucial information for choosing the appropriate face detector. This helps to improve the accuracy and reduce the computation in multi-view face detection compared to other methods. For video sequences, further computational reduction can be achieved by using a pose change smoothing strategy. When face detectors find a face in frontal view, a support vector machine-based multi-class classifier is activated for face recognition. All the above issues are integrated under a support vector machine framework. Test results on four video sequences are presented, among them the detection rate is above 95%, recognition accuracy is above 90%, average pose estimation error is around 10/spl deg/, and the full detection and recognition speed is up to 4 frames/second on a Pentium II 300 PC.

Support vector machine based multi-view face detection and recognition

Detecting faces across multiple views is more challenging than in a fixed view, e.g. frontal view, owing to the significant non-linear variation caused by rotation in depth, self-occlusion and self-shadowing. To address this problem, a novel approach is presented in this paper. The view sphere is separated into several small segments. On each segment, a face detector is constructed. We explicitly estimate the pose of an image regardless of whether or not it is a face. A pose estimator is constructed using Support Vector Regression. The pose information is used to choose the appropriate face detector to determine if it is a face. With this pose-estimation based method, considerable computational efficiency is achieved. Meanwhile, the detection accuracy is also improved since each detector is constructed on a small range of views. We developed a novel algorithm for face detection by combining the Eigenface and SVM methods which performs almost as fast as the Eigenface method but with a significant improved speed. Detailed experimental results are presented in this paper including tuning the parameters of the pose estimators and face detectors, performance evaluation, and applications to video based face detection and frontal-view face recognition. q 2004 Elsevier B.V. All rights reserved.

Modelling faces dynamically across views and over time

A comprehensive novel multi-view dynamic face model is presented in this paper to address two challenging problems in face recognition and facial analysis: modelling faces with large pose variation and modelling faces dynamically in video sequences. The model consists of a sparse 3D shape model learnt from 2D images, a shape-and-pose-free texture model, and an affine geometrical model. Model fitting is performed by optimising (1) a global fitting criterion on the overall face appearance while it changes across views and over time, (2) a local fitting criterion on a set of landmarks, and (3) a temporal fitting criterion between successive frames in a video sequence. By temporally estimating the model parameters over a sequence input, the identity and geometrical information of a face is extracted separately. The former is crucial to face recognition and facial analysis. The latter is used to aid tracking and aligning faces. We demonstrate the results of successfully applying this model on faces with large variation of pose and expression over time.

Multi-view face detection using support vector machines and eigenspace modelling

An approach to multi-view face detection based on head pose estimation is presented in this paper. Support vector regression is employed to solve the problem of pose estimation. Three methods, the eigenface method the support vector machine (SVM) based method, and a combination of the two methods, are investigated. The eigenface method, which seeks to estimate the overall probability distribution of patterns to be recognised, is fast but less accurate because of the overlap of confidence distributions between face and non-face classes. On the other hand, the SVM method, which tries to model the boundary of two classes to be classified is more accurate but slower as the number of support vectors is normally large. The combined method can achieve an improved performance by speeding up the computation and keeping the accuracy to a preset level. It can be used to automatically detect and track faces in face verification and identification systems.

Constructing facial identity surfaces in a nonlinear discriminating space

Recognising face with large pose variation is more challenging than that in a fixed view, e.g. frontal-view, due to the severe non-linearity caused by rotation in depth, self-shading and self-occlusion. To address this problem, a multi-view dynamic face model is designed to extract the shape-and-pose-free facial texture patterns from multi-view face images. Kernel Discriminant Analysis is developed to extract the significant non-linear discriminating features which maximise the between-class variance and minimise the within-class variance. By using the kernel technique, this process is equivalent to a Linear Discriminant Analysis in a high-dimensional feature space which can be solved conveniently. The identity surfaces are then constructed from these non-linear discriminating features. Face recognition can be performed dynamically from an image sequence by matching an object trajectory and model trajectories on the identity surfaces.

Video-based online face recognition using identity surfaces

A multi-view dynamic face model is designed to extract the shape-and-pose-free texture patterns effaces. The model provides a precise correspondence to the task of recognition since the 3D shape information is used to warp the multi-view faces onto the model mean shape in frontal-view. The identity surface of each subject is constructed in a discriminant feature space from a sparse set of face texture patterns, or more practically, from one or more learning sequences containing the face of the subject. Instead of matching templates or estimating multi-modal density functions, face recognition can be performed by computing the pattern distances to the identity surfaces or trajectory distances between the object and model trajectories. Experimental results depict that this approach provides an accurate recognition rate while using trajectory distances achieves a more robust performance since the trajectories encode the spatio-temporal information and contain accumulated evidence about the moving faces in a video input.

Extraction of Tschebysheff Design Data for the Lowpass Dielectric Multilayer

The extraction of design data for the lowpass dielectric multilayer according to Tschebysheff performance is described. The extraction proceeds initially by analogy with electric-circuit design, and can then be given numerical refinement which is also described. Agreement with the Tschebysheff desideratum is satisfactory. The multilayers extracted by this procedure are of fractional thickness, symmetric with regard to their central layers.

Least Squares Method for the Automatic Design of Multi-layers

A simple routine is given for the automatic design of multilayer systems. In the applications described a least-squares criterion for fitting to a specified reflectance function is used; other criteria may be substituted as required. The method is illustrated with examples of anti-reflection, broad-band reflection, achromatic beam-splitter and cut-on systems.

Theoretical determination of the optical constants of weakly absorbing thin films

Methods for determining the optical constants of thin films from photometric measurements of reflectance and transmittance have been known to fail when applied to weakly absorbing films, because of the ill-conditioning of the equations used in the calculation. A method is presented which overcomes this problem by assuming that the refractive index obeys a dispersion formula of the Sellmeier type and then solving the equations simultaneously over a range of wavelengths. This method has been used to determine the optical constants of various materials in the near-ultraviolet and visible regions of the spectrum; results are given for the oxides of thorium, aluminium, zirconium and neodymium, together with neodymium fluoride.

The DAP subroutine library

Abstract The Distributed Array Processor arrived at Queen Mary College with virtually no existing user oriented software apart from the DAP Fortran compiler; this paper will describe the way in which the DAP Support Unit is developing a comprehensive subroutine library. It will include a summary of the decisions reached on organisation and standards relating to documentation, testing and validation, following discussions with NAG Ltd. Two types of library are envisaged: 1. 1. a “trial” library into which users are invited to submit routines for general use, provide suitable machine based documentation is also supplied, and 2. 2. a “standard” fully validated, tested and documented library. A major source of contributions is from members of user groups in specific areas, which reflect the fields of interest of current DAP users. The structure and contents of the trial library will be described; many of the earlier routines submitted were basic utilities for data organisation, but in recent months there has been an increase in the numerical applications, for example, Fourier transforms, spline evaluation, matrix multiplication and inversion, eigenvalue determination, linear systems solvers, random number generators and graphics. Reference will be made to certain interesting parallel processing techniques used in the design of some of these routines.