Vision GPU

From TEILab
Jump to: navigation, search

Project Team

  • Project Lead: Seung In Park 'Graduated 'PhD Student
  • Dr. Francis Quek Professor, Center for Human Computer Interaction
  • Dr. Yong Cao Assistant Professor, Center for Human Computer Interaction
  • Jing Huang Graduated-PhD Student
  • Sean P. Ponce Graduated-MS Student

Note: This is a recently completed project that is currently dormant

Project Statement

The goal of Vision-GPU is to achieve real time image and multimedia processing techniques using modern programmable Graphic Processing Units (GPU).

Project Overview

Real-time image processing on video frames is difficult to attain even with the most powerful modern CPUs. Increasing resolution of video capture devices and increased requirement for accuracy make it is harder to realize real-time performance. To address the high computational demands of video and multimedia data processing, Vision-GPU research explores the computational power from widely available, low-cost Graphics Processing Units (GPUs). We already demonstrated standard image and video analysis operations, such as convolution, edge detection, and motion tracking can have more than 10 times speedup on most of GPUs over single high-end CPUs. Our goal is to show the viability of applying GPUs to general image and multimedia processing, and provide optimization strategies for various vision processing operations using GPUs.

Vcm result2.jpg

Figure 1. Vector Coherence Mapping (VCM) processing results with GPU implementation. VCM algorithm running on NVIDIA 8800GTS-512 showed 22.96 times performance enhancement compared to it's running on Intel 3.2 GHz Pentium 4.


  • Jing Huang, Sean P. Ponce, Seung In Park, Yong Cao, and Francis Quek, GPU-Accelerated Computation for Robust Motion Tracking Using the CUDA Framework, The 5th International Conference on Visual Information Engineering, July, 2008, Xi’an, China.
  • Seung In Park, Sean P. Ponce, Jing Huang, Yong Cao and Francis Quek, Low-Cost, High-Speed Computer Vision Using NVIDIA’s CUDA Architecture, Applied Imagery Pattern Recognition 2008, Dec, 2008, Washington D.C., Virginia, USA.


This research has been partially supported by NSF grants “Embodied Communication: Vivid Interaction with History and Literature,” IIS-0624701, “Interacting with the Embodied Mind,” CRI-0551610, and “Embodiment Awareness, Mathematics Discourse and the Blind,” NSF-IIS- 0451843.