Wavelet frame accelerated reduced vector machine for efficient image analysis

We propose a new approach for face and facial feature detection combined with the advantages of the Morphable Model. The presented method reduces the runtime complexity of a Support Vector Machine classifier and the new training algorithm is fast and simple. This is achieved by an Over-Complete Wavelet Transform that finds optimally sparse approximations of the Support Set Vectors. The wavelet-based approach provides an upper bound on the distance between the decision function of the Support Vector Machine and our classifier. The obtained classifier is fast since the used Haar wavelet approximations of the Support Set Vectors allow efficient Integral Image-based kernel evaluations. This provides a set of double-cascaded classifiers of increasing accuracy for an early rejection. The algorithm yields an excellent runtime performance that is achieved by hierarchically discriminating with respect to the number and approximation accuracy of incorporated Reduced Set Vectors. The proposed algorithm is applied to the problem of face and facial feature detection, but it can also be used for other image-based classifications. The algorithm presented, provides a 530-fold speed-up over the Support Vector Machine, enabling face detection at more than 25 fps on a standard PC. Summarizing, we propose very fast and efficient to train classifiers that improve the detection performance by involving the advantages of the Morphable Model. On one hand to improve the fitting algorithm of the Morphable Model by automatic anchor point detection and on the other hand to use the Morphable Model for improving the training by synthetic data sets and to reduced the False Acceptance Rate.