Recognizing facial expressions (eg. smiling) from different view-points is performed using an Adaboost classifier. In a pre-processing step the positions of the eyes, nose and mouth are detected and used to register both training and test images. A Viola & Jones style classifier is then learned to distinguish between different facial expressions. In further experiments, an Active Contour Model (Snake) is implemented and used to enhance feature extraction.
Face images taken from different view-points
The aim or this project is to reconstruct the original 3D vessel tree using two x-ray projections of the coronary vein. We approached this problem using a stochastic Branch & Bound method to identify corresponding vessel segments in the two images. With this correspondence information, and knowing the position and parameters of the imaging system involved, reconstruction of the coronary tree can be performed using simple ray intersection
Corresponding vessel bifurcations in two x-ray projections
This project is about recognising biological cells in water. Our approach is based on a nearest-neighbour similarity matching algorithm which uses multi-scale Gaussian differential features. Experiments are preformed on two sets of images containing: (i) 7 different cell shape classes (e.g. round, tear-drop, flagella) and (ii) 6 different cell types (e.g. c. debile, n. closterium, t. dicipiens).
Examples of different cell types