Grant: Engage Grants Program
École Polytechnique de Montréal
$25,000.00
Optimization of a simple lens deblurring algorithm for graphics processing units
FY 2014-15
Summary
Algolux is a startup established in Montréal in 2013. Supported by Tandem Lauch and several other partners,with a staff of 11 people, Algolux embraces ever-increasing computational power to improve optics in compactportable devices. Algolux uses computational photography to fix blurring problems traditionally solved by thephysics of optics. This means that less optical elements are needed to correct aberrations, resulting in clearerpictures from thinner and lighter camera modules at lower manufacturing costs. This is challenging as thecomputational power and available energy are very limited in portable devices. Algolux is thus pursuing theoptimization of its proprietary method so that it can eventually run in real time for videos (30fps for 2 Mpixelimages). The proposal is to parallelize part of the algorithm on Graphics Processing Units (GPU). GPUs areused as an alternative to central processing units (CPU) for computation intensive and throughput orientedalgorithms. Typical GPUs have a single instruction, multiple data (SIMD) architecture which takes advantageof its hundred to thousand cores. Two aspects will be considered in this project: how to spread the computationamong computation threads in an effective way and how to create thread exploiting instruction-levelparallelism. It is expected that GPUs will become a standard feature of smartphones and tablets in the nearfuture. Hence there is a pressing need to adapt the simple lens algorithm for its deployment in mobile devices.The general goal of this project is to refine and optimize existing algorithms and techniques to achieve and thusdemonstrate real time execution of simple lens deblurring in a GPU-equipped mobile device. The project isdivided in three major phases. First, a profiling will be done on the algorithm to extract the parallelizablesections. Then, anticipated GPU architectures will be studied to find the most suitable one for the simple lensdeblurring algorithm. Finally, the implementation will be made on GPU taking into account two importantfactors: computations and memory accesses to fully optimize the algorithms.
Program
Engage Grants Program
Recipient
Location
Quebec, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
469910-2014
Name nom
Savaria, Yvon
Department département
Génie électrique
Organization id
64
Institution établissement
École Polytechnique de Montréal
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2014
Competition year année de concours
2014
Award amount
25000
Program id
EGP
Program name en
Engage Grants Program
Program name fr
Programme de subventions d'engagement partenariat
Group en
Research partnerships programs
Group fr
Programmes de partenariat
Committee code
1552
Committee name en
Quebec Internal Decision Committee
Committee name fr
Comité de décision interne pour le Québec
Area of application code
801
Area of application group en
Information and communication services
Area of application group fr
Services d'information et de télécommunications
Area of application en
Computer software
Area of application fr
Logiciel
Research subject code
2708
Research subject group en
Information technology
Research subject group fr
Technologie de l'information
Research subject en
Image processing
Research subject fr
Traitement des images
Partie
1
Nb partie
1
Application title
Optimization of a simple lens deblurring algorithm for graphics processing units
Keyword
NSERC
Application summary
Algolux is a startup established in Montréal in 2013. Supported by Tandem Lauch and several other partners,with a staff of 11 people, Algolux embraces ever-increasing computational power to improve optics in compactportable devices. Algolux uses computational photography to fix blurring problems traditionally solved by thephysics of optics. This means that less optical elements are needed to correct aberrations, resulting in clearerpictures from thinner and lighter camera modules at lower manufacturing costs. This is challenging as thecomputational power and available energy are very limited in portable devices. Algolux is thus pursuing theoptimization of its proprietary method so that it can eventually run in real time for videos (30fps for 2 Mpixelimages). The proposal is to parallelize part of the algorithm on Graphics Processing Units (GPU). GPUs areused as an alternative to central processing units (CPU) for computation intensive and throughput orientedalgorithms. Typical GPUs have a single instruction, multiple data (SIMD) architecture which takes advantageof its hundred to thousand cores. Two aspects will be considered in this project: how to spread the computationamong computation threads in an effective way and how to create thread exploiting instruction-levelparallelism. It is expected that GPUs will become a standard feature of smartphones and tablets in the nearfuture. Hence there is a pressing need to adapt the simple lens algorithm for its deployment in mobile devices.The general goal of this project is to refine and optimize existing algorithms and techniques to achieve and thusdemonstrate real time execution of simple lens deblurring in a GPU-equipped mobile device. The project isdivided in three major phases. First, a profiling will be done on the algorithm to extract the parallelizablesections. Then, anticipated GPU architectures will be studied to find the most suitable one for the simple lensdeblurring algorithm. Finally, the implementation will be made on GPU taking into account two importantfactors: computations and memory accesses to fully optimize the algorithms.