< Back to previous page
Project
A combined GP-GPU/FPGA desktop system for accelerating image processing application (GUDI). (IWT514)
This projects targets companies which are in need of more processing power for executing their applications in a reasonable time. The project focuses on image processing in the medical sector, but the generic results will be widely applicable. The project will offer companies the opportunity to learn in an efficient and cost-effective way the differences and the programming techniques of GPUs and FPGAs, and lower the threshold for using them.
Besides new and more complex algorithms, we are facing an increase in data, by for example an increase in image resolution and faster acquisition techniques. The required computational resources can, however, not be offered since Moore's law has stopped for single-core processors. Alternative solutions have to be considered to offer the necessary processing power for applying the current algorithms on the new data volumes.
Currently, there are different special purpose processors on the market that are designed to speed up these compute-intensive algorithms. Examples of such accelerators that can often achieve better performance than CPUs on certain workloads are Field Programmable Gate Arays (FPGAs), Graphical Processing units (GPUs), the Cell Broadband Engine Cell/B.E., Digital Signal Processors (DSPs), network processors,... We will focus on the first 2 types.
These new technologies offer opportunities to tackle both problems. For the moment, developers are facing the problem that it is not clear in which technology to invest. This project will attempt to acquire the necessary knowhow.
In a first step, we will investigate the performance between the different technologies by means of certain test cases of algorithms in the sector of image processing. Most of these algorithms are expected to possess major acceleration on the GPU since many image processing techniques consist of a common computation over all the pixels. Parameters such as speed, power consumption, programming effort, maintainability, cost, chip area, accuracy will all be taken into account.
Secondly, we will create a library for a selected set of image processing algorithms in the different technologies. This will facilitate companies with the introduction of a new technology in their applications and also accelerate the learning process of a new programming language.
Besides new and more complex algorithms, we are facing an increase in data, by for example an increase in image resolution and faster acquisition techniques. The required computational resources can, however, not be offered since Moore's law has stopped for single-core processors. Alternative solutions have to be considered to offer the necessary processing power for applying the current algorithms on the new data volumes.
Currently, there are different special purpose processors on the market that are designed to speed up these compute-intensive algorithms. Examples of such accelerators that can often achieve better performance than CPUs on certain workloads are Field Programmable Gate Arays (FPGAs), Graphical Processing units (GPUs), the Cell Broadband Engine Cell/B.E., Digital Signal Processors (DSPs), network processors,... We will focus on the first 2 types.
These new technologies offer opportunities to tackle both problems. For the moment, developers are facing the problem that it is not clear in which technology to invest. This project will attempt to acquire the necessary knowhow.
In a first step, we will investigate the performance between the different technologies by means of certain test cases of algorithms in the sector of image processing. Most of these algorithms are expected to possess major acceleration on the GPU since many image processing techniques consist of a common computation over all the pixels. Parameters such as speed, power consumption, programming effort, maintainability, cost, chip area, accuracy will all be taken into account.
Secondly, we will create a library for a selected set of image processing algorithms in the different technologies. This will facilitate companies with the introduction of a new technology in their applications and also accelerate the learning process of a new programming language.
Date:1 Jan 2011 → 28 Feb 2013
Keywords:Low Power Cmos, Numerical Linear Algebra, Digital Image Processing, Image Reconstruction, Embedded System Design, Displays, Audio Processing, Light Detectors, Micro-Electronics Technology, Sige Bicmos Design, Satellite Image Analysis, Medical Image Analysis, Inverse Problems, Telemedicine, JPEGx, Video Compression, Neural Networks, Mine Detection, Vision, Digital Signal Processing, Electronic System Design, Machine Vision, Micro Electronics, Chip Interconnects (Inter / Intra), Cmos Design, Humanitarian Demining, Speech Processing, Mpegx, Light Emitters, Pattern Recognition, Mm-Wave Technology, Robot Vision, Impedance Tomography, Image Compression, Light Modulators, Medical Image Visualization, Opto-Electronics, Motion Estimation And Tracking, Computer Aided Electronic Design, Multispectral Image Analysis, Computer Vision, Image Processing, Industrial Visual Inspection, Image Analysis, electronics
Disciplines:Mathematical sciences and statistics, Electrical and electronic engineering, (Bio)medical engineering