3D reconstruction by a combined structure tensor and Hough transform light field approach
-
Alessandro Vianello
Alessandro Vianello received his Bachelor of Science in Telecommunications Engineering in 2010, and his Master of Science in Telecommunications Engineering in 2013 from the University of Padova. Since 2014 he is pursuing his PhD in Computer Science at the University of Heidelberg, under the supervision of Prof. Bernd Jähne and funded by Robert Bosch GmbH. His research interests include image processing, depth estimation, 3D geometry reconstruction, and sensor fusion.
Robert Bosch GmbH, Robert Bosch Campus 1, 71272 Renningen, Germany
,
Giulio Manfredi
Giulio Manfredi received his Bachelor of Science in Cinema and Media Engineering from the Polytechnic University of Turin and his Master of Science in Applied Computer Science from the University of Heidelberg with a thesis on depth estimation from 3D light fields.
Heidelberg University, Heidelberg Collaboratory for Image Processing (HCI), Berliner Straße 43, 69120 Heidelberg, Germany
Maximilian Diebold is leader of the Light-Field Imaging Group at the Heidelberg Collaboratory for Image Processing since April 2016. He studied Electrical Engineering and Information Technology at the University of Karlsruhe (Dipl.-Ing.) and received his PhD in computer science at the combined faculty for the natural science and mathematics at the University of Heidelberg (Dr.rer.nat.) in April 2016. The thesis was collaborated with Sony Germany GmbH. His current interest is to decompose image information by taking a holistic point of view about depth, material, reflectance, color, polarization and illumination from a large set of optical measurements i.e. light-field measurements and use this information to generate what he terms tunable ground truth.
Heidelberg University, Heidelberg Collaboratory for Image Processing (HCI), Berliner Straße 43, 69120 Heidelberg, Germany
Bernd Jähne received his Diploma, Doctoral degree and Habilitation degree in Physics from Heidelberg University in 1977, 1980, and 1985, respectively, and a Habilitation degree in Applied Computer Science from the University of Hamburg-Harburg in 1992. From 1988 to 2003 he hold a research professorship at the Scripps Institution of Oceanography, University of California in San Diego. Since 1994 he is professor at the Interdisciplinary Center for Scientific Computing (IWR) and Institute for Environmental Physics of Heidelberg University in 1994 and since 2008 he heads the Heidelberg Collaboratory for Image Processing (HCI). His research interests include small-scale air-sea interaction and image processing.
Heidelberg University, Heidelberg Collaboratory for Image Processing (HCI), Berliner Straße 43, 69120 Heidelberg, Germany
Abstract
Disparity estimation using the structure tensor is a local approach to determine orientation in Epipolar Plane Images. A global extension would lead to more precise and robust estimations. In this work, a novel algorithm for 3D reconstruction from linear light fields is proposed. This method uses a modified version of the Progressive Probabilistic Hough Transform to extract orientations from Epipolar Plane Images, allowing to achieve high quality disparity maps. To this aim, the structure tensor estimates are used to speed up computation and improve the disparity estimation near occlusion boundaries. The new algorithm is evaluated on both synthetic and real light field datasets, and compared with classical local disparity estimation techniques based on the structure tensor.
Zusammenfassung
Die Disparitätsschätzung mittels eines Strukturtensors ist ein lokaler Ansatz zur Bestimmung der Orientierung in Epipolar Plane Images. Eine Erweiterung zum Globalen würde präzisere und robustere Schätzungen ermöglichen. In dieser Arbeit wird ein neuartiger Algorithmus zur 3D-Rekonstruktion aus linearen Lichtfeldern vorgeschlagen. Er verwendet eine modifizierte Version der Progressive Hough-Transformation in Kombination mit dem Strukturtensor, um Orientierungen aus Kantenbildern der Epipolar Plane Images zu extrahieren und dadurch eine hohe Qualität in den Disparitätskarten zu erreichen. Dabei werden die Schätzwerte des Strukturtensors benutzt, um die Berechnungen zu beschleunigen und um die Disparitätsschätzung in der Nähe von Verdeckungskanten zu verbessern. Der neue Algorithmus wird sowohl auf synthetischen als auch auf echten Lichtfelddatensätzen evaluiert und mit klassischen, lokalen Schätzmethoden basierend auf dem Strukturtensor verglichen.
About the authors
Alessandro Vianello received his Bachelor of Science in Telecommunications Engineering in 2010, and his Master of Science in Telecommunications Engineering in 2013 from the University of Padova. Since 2014 he is pursuing his PhD in Computer Science at the University of Heidelberg, under the supervision of Prof. Bernd Jähne and funded by Robert Bosch GmbH. His research interests include image processing, depth estimation, 3D geometry reconstruction, and sensor fusion.
Robert Bosch GmbH, Robert Bosch Campus 1, 71272 Renningen, Germany
Giulio Manfredi received his Bachelor of Science in Cinema and Media Engineering from the Polytechnic University of Turin and his Master of Science in Applied Computer Science from the University of Heidelberg with a thesis on depth estimation from 3D light fields.
Heidelberg University, Heidelberg Collaboratory for Image Processing (HCI), Berliner Straße 43, 69120 Heidelberg, Germany
Maximilian Diebold is leader of the Light-Field Imaging Group at the Heidelberg Collaboratory for Image Processing since April 2016. He studied Electrical Engineering and Information Technology at the University of Karlsruhe (Dipl.-Ing.) and received his PhD in computer science at the combined faculty for the natural science and mathematics at the University of Heidelberg (Dr.rer.nat.) in April 2016. The thesis was collaborated with Sony Germany GmbH. His current interest is to decompose image information by taking a holistic point of view about depth, material, reflectance, color, polarization and illumination from a large set of optical measurements i.e. light-field measurements and use this information to generate what he terms tunable ground truth.
Heidelberg University, Heidelberg Collaboratory for Image Processing (HCI), Berliner Straße 43, 69120 Heidelberg, Germany
Bernd Jähne received his Diploma, Doctoral degree and Habilitation degree in Physics from Heidelberg University in 1977, 1980, and 1985, respectively, and a Habilitation degree in Applied Computer Science from the University of Hamburg-Harburg in 1992. From 1988 to 2003 he hold a research professorship at the Scripps Institution of Oceanography, University of California in San Diego. Since 1994 he is professor at the Interdisciplinary Center for Scientific Computing (IWR) and Institute for Environmental Physics of Heidelberg University in 1994 and since 2008 he heads the Heidelberg Collaboratory for Image Processing (HCI). His research interests include small-scale air-sea interaction and image processing.
Heidelberg University, Heidelberg Collaboratory for Image Processing (HCI), Berliner Straße 43, 69120 Heidelberg, Germany
Acknowledgement
The authors gratefully acknowledge financial support for this research by the Robert Bosch GmbH (PhD scholarship for Alessandro Vianello) and the Heidelberg Collaboratory for Image Processing (HCI) within the Institutional Strategy ZUK49 “Heidelberg: Realizing the Potential of a Comprehensive University”, Measure 6.4 including matching funds from the industry partners of the HCI.
©2017 Walter de Gruyter Berlin/Boston
Articles in the same Issue
- Frontmatter
- Editorial
- Forum Bildverarbeitung 2016
- Beiträge
- On the design of a fractal calibration pattern for improved camera calibration
- Compressive shape from focus based on a linear measurement model
- 3D reconstruction by a combined structure tensor and Hough transform light field approach
- Das baryzentrische Verhältnis als Invariante der projektiven Geometrie
- Simulation of microscopic metal surfaces based on measured microgeometry
- Automated surface inspection of small customer-specific optical elements
- Individualisierte optische Messtechnik basierend auf additiv gefertigten optischen Komponenten
- Monokulare Kopfposenschätzung basierend auf dem optischen Fluss und der Verfolgung stabiler Merkmale
- Ein intuitives kamerabasiertes System zur Assistenz sehbehinderter Menschen
Articles in the same Issue
- Frontmatter
- Editorial
- Forum Bildverarbeitung 2016
- Beiträge
- On the design of a fractal calibration pattern for improved camera calibration
- Compressive shape from focus based on a linear measurement model
- 3D reconstruction by a combined structure tensor and Hough transform light field approach
- Das baryzentrische Verhältnis als Invariante der projektiven Geometrie
- Simulation of microscopic metal surfaces based on measured microgeometry
- Automated surface inspection of small customer-specific optical elements
- Individualisierte optische Messtechnik basierend auf additiv gefertigten optischen Komponenten
- Monokulare Kopfposenschätzung basierend auf dem optischen Fluss und der Verfolgung stabiler Merkmale
- Ein intuitives kamerabasiertes System zur Assistenz sehbehinderter Menschen
