Neural mechanisms of feature extraction for the analysis of shape and behavioral patterns [Elektronische Ressource] / vorgelegt von Ulrich Weidenbacher

NeuralmechanismsoffeatureextractionfortheanalysisofshapeandbehavioralpatternsDissertation zur Erlangung des Doktorgrades (Dr.rer.nat.)der Fakultät für Informatik und Ingenieurwissenschaftender Universität Ulm,vorgelegt vonUlrich Weidenbacheraus Heidenheim a.d. Brenz2010Universität UlmInstitut für NeuroinformatikAmtierender Dekan: Prof. Dr. Michael WeberErstgutachter: Prof. Dr. Heiko NeumannZweitgutachter: Prof. Dr. Günther PalmTag der Promotion: 7.12.2010AbstractThehumanvisualsystemsegments3Dscenesinsurfacesandobjectswhichcanappearat different depths with respect to the observer. The projection from 3D to 2D leads par-tially to occlusions of objects depending on their position in depth. There is experimentalevidence that surface-based features such as occluding contours or junctions are used ascues for the robust segmentation of surfaces. These features are characterized by theirrobustness against variations of illumination and small changes in viewpoint. We demon-strate that this feature representation can be used to extract a sketch-like representationof salient features that captures and emphasizes perceptually relevant regions on objectsand surfaces. Furthermore, this representation is also suitable for learning more complexform patterns such as faces and bodies in different pose.In this thesis, we present a biologically inspired, recurrent model which extracts andinterprets surface-based features from a 2D grayscale intensity input image.