• COMPUTER VISION
• VISUAL ATTENTION

The thesis presents an algorithm for object detection based on a computational model of visual attention. The algorithm learns the description of a new object using a bottom-up, stimulus-driven approach to attention. Once an object has been learnt, the algorithm uses the object description in order to perform a top-down attention-guided search looking for candidate object locations. The search takes into account differences in scale and rotations between the image and the object prototype.
Furthermore, the thesis describes an innovative single-modality image registration algorithm based on Swarm Intelligence principles and experimental work on medical images.