Resumen de Simultaneous detection and segmentation for generic objects
Albert Torrent Palomeras
This thesis deals with the simultaneous detection and segmentation for generic objects in images. The proposed approach is based on building a dictionary of patches, which defines the object and allows the extraction of the detection and segmentation features. These features are then used in a boosting classifier which automatically decides at each round whether it is better to detect or segment. Moreover, we include in the boosting training the ability of crossing information between detection and segmentation with the aim that good detections may help to better segment and vice versa. The experimental results obtained using three different datasets show a good performance both in detection and segmentation, with results comparable to state of the art approaches.
We also propose to use our simultaneous detection and segmentation approach to automatically annotate images downloaded using Internet search engines, providing polygonal annotations of the objects. The system only requires the user feedback for validating the automatic annotations provided by the classifiers.
Finally, we adapt also the detection proposal to deal with specific problems of object recognition in different areas. On the one hand, we present a new fully automatic computer aided detection system for microcalcification detection. On the other hand, a novel approach for the detection of faint compact sources in wide field interferometric radio images has been proposed. The results obtained in both cases demonstrate the validity of using our approach in such specific problems with simple modifications. This point stresses one of the objectives of this thesis; proposing a generic approach able to deal with objects of a very different nature.
