|
Dr J. Paul Siebert
BSc, PhD, MIET, CEng. , MIEEE
I am a Reader in the Department of Computing Science and leader of the Computer Vision & Graphics Group.
Announcements
An Introduction to 3D Computer Vision Techniques and Algorithms, by Bogusław Cyganek & J. Paul Siebert, Published Wiley, January 2009, now in second edition. Finalist in the 2009 Thales Scottish Technology Competition. More details: New Electronics 2008 Medical Futures Award, in a team led by Prof Gordon Dutton that developed the SightSim TM system which enables parents to see the world as their visually-impaired child does. More details: GU News, BBC News, STV News, Scotland’s Oracle Research Interests
Keywords: Computer vision; vision science; image processing; metric 3D imaging; 3D imaging applications: whole human body 3D capture and anatomy modelling, skin imaging (visible and infrared), skin modelling, soft tissue modelling, 3D surgery planning tools, forensic facial reconstruction, realistic virtual actors & personal avatars. Robot vision; anthropomorphic vision approaches; active vision & visuo-motor behaviours; space variant imaging (foveated vision), Retino-Cortical transform (complex-log & log-polar transforms); visual attention, scale-space representations and appearance-based imaging, SIFT 2.5D SIFT, local feature representations, cognitive vision.
Clinical 3D Imaging: My principal area of research has centred on developing photogrammetry-based 3D imaging systems for clinical, media and robotics applications. Clinical research, in collaboration with the Department of Statistics and Glasgow Dental Hospital, includes the development of tools for assessing 3D human (and animal) surface anatomy. My current focus is the development of an entirely new approach to representing, manipulating and analysing 3D data that exploits 2.5D range maps directly, without the need for intermediate polygonisation and its associated computational inefficiencies. To this end we have extended the SIFT transform into the 2.5D domain and several of my PhD students are investigating related topics in this area. I am currently a co-investigator in Wellcome Trust funded project led by the Department of Statistics that seeks to automate fully the measurement and analysis of 3D face scans.
Media Related Research: I have been involved in a number of media-related projects, for example, 3D whole body imaging (SHEFC RDG), developing 3D imaging approaches based on high resolution passive capture that does not require active illumination (EPSRC) but can provide very high quality photorealistic 3D models (EPSRC). The most significant media work I have supervised is in human character animation based on the conformation of a generic deformable mesh to 3D scan data (EPSRC). This work was developed to pre-production quality in collaboration with Edinburgh University Informatics Division to demonstrate the concept of cloning 3D scanned individuals such that they can be animated within virtual media productions (Scottish Enterprise).
Robot Vision: From the outset of my research in 3D imaging I have been interested in active binocular vision systems, and have investigated semi-autonomous 3D vision systems to provide automatic vergence control for telepresence systems and systems capable of automatic 3D scene recovery. In this latter case, a foveated imaging approach has been adopted in combination with a gaze control strategy based on detecting interest points on the foveated 2.5D range surface topology (EPSRC Industrial CASE). A visual search strategy driven by these interest points is used to cause the binocular camera system to saccade and thereby explore the scene, capturing 2.5D range maps which are integrated into a 3D model on the fly. I am currently supervising ongoing research work in robot vision that adopts a biologically motivated, space variant vision model, currently based on a randomly sampled retina. We have now constructed a complete foveated vision system that is being applied to indexing and annotation of digital moving image sequences, though this system is equally well suited to binocular robot vision applications. A longstanding goal in the vision community is to build a robot vision system that is capable of autonomously exploring its environment and a binocular robot head has been developed in CV&G that can find known objects under cluttered and partially occluded conditions with a high degree of reliability. Our ongoing research seeks to extend the visual descriptors used in this system, currently based on SIFT, achieve automated object model learning and couple our hierarchy of visual behaviours driving gaze control to a cognitive reasoning system.
Selected Recent Publications
Unsupervised clustering in Hough space for recognition of multiple instances of the same object in a cluttered scene Local Feature Extraction and Matching on Range Images: 2.5D SIFT Self-Correction of 3D Reconstruction from Multi-view Stereo Images Towards a Unified Visual Framework in a Binocular Active Robot Vision System SIFT Keypoint Descriptors for Range Image Analysis Iconic Object-based Saccade Generation using a Biologically Inspired Self-organized Retina Labelling of Three Dimensional Human Body Scans: A Topological Approach Towards building a photo-realistic virtual human face for craniomaxillofacial diagnosis and treatment planning Hierarchical Feature Extraction using a Self-organised Retinal Receptive Field Sampling Tessellation A functional-based segmentation of human body scans in arbitrary postures
Current & Recent Projects
Face 3D: A multi-centre 3 year project carrying out research into the analysis of three-dimensional facial dysmorphology funded by the Wellcome Trust. IP-RACINE, EC integrated project on the Digital Film production chain from capture to playout, UPF Barcelona coordinating partner. CLEFT 10, A multidisciplinary assessment of residual deformities following surgical repair of cleft lip and palate, Scottish Executive Health Department funded, in collaboration with Glasgow Dental School (GU, lead partner), Department of Statistics (GU) and the Department of Psychology (GU)
Contact:
J. Paul Siebert Department of Computing Science, Lillybank Gardens, University of Glasgow, G12 8QQ Scotland UK
Phone:+44 (0) 141 330 3124 Fax: +44(0) 141 330 3119 Email: psiebert@dcs.gla.ac.uk
|
