Toward a Multi-level Parallel Framework on GPU Cluster with PetSC-CUDA for PDE-based Optical Flow Computation

In this work we present a multi-level parallel framework for the Optical Flow computation on a GPUs cluster, equipped with a scientific computing middleware (the PetSc library). Starting from a flow-driven isotropic method, which models the optical flow problem through a parabolic partial differential equation (PDE), we have designed a parallel algorithm and its software implementation that is suitable for heterogeneous computing environments (multiprocessor, single GPU and cluster of GPUs). The proposed software has been tested on real SAR images sequences. Numerical experiments highlight the performance of the proposed software framework, which can reach a gain of about 95% with respect to the sequential implementation.

[1]  Y. Saad,et al.  GMRES: a generalized minimal residual algorithm for solving nonsymmetric linear systems , 1986 .

[2]  David J. Fleet,et al.  Performance of optical flow techniques , 1994, International Journal of Computer Vision.

[3]  Li Song,et al.  Paralleling variable block size motion estimation of HEVC on CPU plus GPU platform , 2013, 2013 IEEE International Conference on Multimedia and Expo Workshops (ICMEW).

[4]  Berthold K. P. Horn,et al.  Determining Optical Flow , 1981, Other Conferences.

[5]  Livia Marcellino,et al.  A numerical algorithm for image sequence inpainting that preserves fine textures , 2011, Int. J. Comput. Math..

[6]  Salvatore Cuomo,et al.  A Smart GPU Implementation of an Elliptic Kernel for an Ocean Global Circulation Model , 2013 .

[7]  Almerico Murli,et al.  Real time ultrasound image sequence segmentation on multicores , 2009, International Conference on Parallel Computing.

[8]  Almerico Murli,et al.  Numerical Solution of Diffusion Models in Biomedical Imaging on Multicore Processors , 2011, Int. J. Biomed. Imaging.

[9]  Almerico Murli,et al.  A Motion-Aided Ultrasound Image Sequence Segmentation , 2010 .

[10]  P. Knabner,et al.  Numerical Methods for Elliptic and Parabolic Partial Differential Equations , 2003, Texts in Applied Mathematics.

[11]  Pablo Montero,et al.  Fast GPU approximation of EPZS motion estimation using branching , 2013, 2013 IEEE 15th International Workshop on Multimedia Signal Processing (MMSP).

[12]  Salvatore Cuomo,et al.  3D Data Denoising via Nonlocal Means Filter by Using Parallel GPU Strategies , 2014, Comput. Math. Methods Medicine.

[13]  Joachim Weickert,et al.  Recursive Separable Schemes for Nonlinear Diffusion Filters , 1997, Scale-Space.

[14]  Salvatore Cuomo,et al.  Surface reconstruction from scattered point via RBF interpolation on GPU , 2013, 2013 Federated Conference on Computer Science and Information Systems.

[15]  T. Brox,et al.  Universität Des Saarlandes Fachrichtung 6.1 – Mathematik a Survey on Variational Optic Flow Methods for Small Displacements a Survey on Variational Optic Flow Methods for Small Displacements , 2022 .

[16]  Antonio García Dopico,et al.  Parallel Computation of Optical Flow , 2004, ICIAR.

[17]  Timo Kohlberger,et al.  Parallel Variational Motion Estimation by Domain Decomposition and Cluster Computing , 2004, ECCV.