Abstract-In this paper, we introduce concepts to reduce the computational complexity of regression, which are successfully used for Support Vector Machines. To the best of our knowledge, we are the first to publish the use of a cascaded Reduced Set Vector approach for regression. The WaveletApproximated Reduced Vector Machine classifiers for face and facial feature point detection are extended to regression for efficient and robust head pose estimation. We use synthetic data, generated by the 3D Morphable Model, for optimal training sets and demonstrate results superior to state-ofthe-art techniques. The new Wavelet Reduced Vector Regression shows similarly good results on natural data, gaining a reduction of the complexity by a factor of up to 560. The introduced Evolutionary Regression Tree uses coarse-tofine loops of strongly reduced regression and classification up to most accurate complex machines. We demonstrate the Cascaded Condensation Tracking for head pose estimation for a large pose range up to ±90 degrees on videostreams.

No abstract

The Karhunen-Loeve Transform (KLT) can be used to compress sets of correlated visual data. Human faces and object recognition are popular areas of current research that use KLT-based methods. The KLT can also be used to compress visual data corresponding to a camera moved translationally and/or rotationally relative to a scene. Positioning of a camera relative to a scene can then be derived accurately using KLT feature vectors; this finds application in robotics and autonomous navigation. Various factors affect the accuracy and speed of such position determination including the number of KLT vectors used, the number of images used to perform the KLT, the number of images used in the comparison set and the size of the movement range. This paper investigates the performance of the KLT with a series of experiments determining a camera's rotational position relative to a generic laboratory scene.

No abstract