This title appears in the Scientific Report :
2014
Please use the identifier:
http://dx.doi.org/10.1007/978-3-642-54420-0_21 in citations.
Computation of Mutual Information Metric for Image Registration on Multiple GPUs
Computation of Mutual Information Metric for Image Registration on Multiple GPUs
Because of their computational power, GPUs are widely used in the field of image processing. Registration of brain images has already been successfully accelerated with GPUs, but registration of high-resolution human brain images presents new challenges due to large amounts of data and images not fi...
Saved in:
Personal Name(s): | Adinets, Andrey (Corresponding Author) |
---|---|
Kraus, Jiri / Axer, Markus / Huysegoms, Marcel / Köhnen, Stefan / Pleiter, Dirk | |
Contributing Institute: |
Strukturelle und funktionelle Organisation des Gehirns; INM-1 Jülich Supercomputing Center; JSC |
Published in: |
Euro-Par 2013: Parallel Processing Workshops |
Imprint: |
Berlin, Heidelberg
Springer Berlin Heidelberg
2014
|
Physical Description: |
208 - 217 |
ISBN: |
978-3-642-54419-4 (print) 978-3-642-54420-0 (electronic) |
DOI: |
10.1007/978-3-642-54420-0_21 |
Conference: | Euro-Par 2013: Parallel Processing Workshops, Aachen (Germany), 2013-08-25 - 2013-08-30 |
Document Type: |
Contribution to a book Contribution to a conference proceedings |
Research Program: |
Supercomputing and Modelling for the Human Brain Supercomputer Facility |
Series Title: |
Lecture Notes in Computer Science
8374 |
Publikationsportal JuSER |
Because of their computational power, GPUs are widely used in the field of image processing. Registration of brain images has already been successfully accelerated with GPUs, but registration of high-resolution human brain images presents new challenges due to large amounts of data and images not fitting in the memory of a single device.In this paper, we address this issue with two approaches. The first approach replicates image data in system memory of each node and distributes only a part of the data over multiple GPUs. The second approach splits image data between multiple GPUs, and overlaps computation and communication to hide latency. For both approaches, we present a performance analysis and comparison. |