OpenCL implementation of basic operations for 3D structure from motion estimation

Structure from Motion (SfM) technique is usually used for camera motion recovery and 3D shape estimation The major problem with most SfM techniques is the computation time. Typically, most SfM techniques involve basic matrix operations. In case of large matrices these operations can increase the computation time of SfM algorithm. Though newer and faster algorithms are developed, as the data increases exponentially, computational processing has been degraded. The trend on computational resources is mostly towards parallel processing. This paper presents OpenCL implementation of a basic matrix operations to optimize computation time of SfM OpenCl is a technique used for parallel computing, which is a form of computation, in which many calculations are carried out simultaneously. OpenCL implementation is compared to sequential implementation and experimental results shows that OpenCL implementation runs several time faster than sequential one.

[1]  Luc Soler,et al.  Fast 3D Structure From Motion with Missing Points from Registration of Partial Reconstructions , 2012, AMDO.

[2]  David R. Kaeli,et al.  Heterogeneous Computing with OpenCL - Revised OpenCL 1.2 Edition , 2012 .

[3]  Hauke Strasdat,et al.  Real-time monocular SLAM: Why filter? , 2010, 2010 IEEE International Conference on Robotics and Automation.

[4]  Tobias Höllerer,et al.  Evaluation of Interest Point Detectors and Feature Descriptors for Visual Tracking , 2011, International Journal of Computer Vision.

[5]  Bernhard P. Wrobel,et al.  Multiple View Geometry in Computer Vision , 2001 .

[6]  Chao Huang,et al.  Accelerating image reconstruction in three-dimensional optoacoustic tomography on graphics processing units. , 2013, Medical physics.

[7]  Takeo Kanade,et al.  A sequential factorization method for recovering shape and motion from image streams , 1997, IEEE Trans. Pattern Anal. Mach. Intell..

[8]  David Kaeli,et al.  Heterogeneous Computing with OpenCL , 2011 .

[9]  Luc Soler,et al.  Evaluation of Endoscopic Image Enhancement for Feature Tracking: A New Validation Framework , 2013, AE-CAI.

[10]  Sinisa Segvic,et al.  Online/Realtime Structure and Motion for General Camera Models , 2008, 2008 IEEE Workshop on Applications of Computer Vision.

[11]  Axel Pinz,et al.  Fast and Globally Convergent Structure and Motion Estimation for General Camera Models , 2006, BMVC.

[12]  Takeo Kanade,et al.  Shape and motion from image streams under orthography: a factorization method , 1992, International Journal of Computer Vision.

[13]  Antoni Grau,et al.  Monocular SLAM for Visual Odometry: A Full Approach to the Delayed Inverse-Depth Feature Initialization Method , 2012 .

[14]  Riccardo Mazzon.  Real-time structure from motion for monocular and stereo cameras , 2010, Melecon 2010 - 2010 15th IEEE Mediterranean Electrotechnical Conference.

[15]  P. J. Narayanan,et al.  Practical Time Bundle Adjustment for 3D Reconstruction on the GPU , 2010, ECCV Workshops.

[16]  Chia-Hsiang Wu,et al.  Three-Dimensional Modeling From Endoscopic Video Using Geometric Constraints Via Feature Positioning , 2007, IEEE Transactions on Biomedical Engineering.

[17]  Ethan Eade,et al.  Monocular Simultaneous Localisation and Mapping , 2008 .

[18]  Dan C. Marinescu,et al.  Parallel algorithms for 3D reconstruction of asymmetric objects from electron micrographs , 1999, Proceedings 13th International Parallel Processing Symposium and 10th Symposium on Parallel and Distributed Processing. IPPS/SPDP 1999.

[19]  Michel Dhome,et al.  Generic and real-time structure from motion using local bundle adjustment , 2009, Image Vis. Comput..