Title :
GPU-based face tracking at 500 fps
Author :
Ishii, Idaku ; Ichida, Hiroki ; Takaki, Takeshi
Author_Institution :
Hiroshima Univ., Hiroshima, Japan
Abstract :
Real-time face tracking at 500 fps was realized for an 8-bit 512×512 color image on a GPU-based high-speed vision platform by implementing a boosting-based face detection algorithm. This algorithm was developed by enhancing the Viola-Jones face detector. In the implementation, the size and position of a face pattern in an image can be efficiently extracted at high speed by reducing the number of window searches for Haar-like features and by combining skin color extraction with Haar classifiers, while assuming a property of high-frame-rate (HFR) vision, i.e., a small interframe image displacement. Several experiments were performed by using rapidly moving face patterns in order to verify the effectiveness of the proposed face tracking system.
Keywords :
Haar transforms; face recognition; feature extraction; graphics processing units; image classification; image colour analysis; image enhancement; image motion analysis; GPU based face tracking; GPU based high-speed vision platform; Haar classifiers; Haar-like features; Viola-Jones face detector; boosting based face detection algorithm; color image; high frame rate vision; image enhancement; interframe image displacement; moving face pattern; skin color extraction; window searches; Cameras; Conferences; Head; Image processing; Haar-like features; High-speed vision; face detection; skin color extraction; target tracking;
Conference_Titel :
Image Processing (ICIP), 2011 18th IEEE International Conference on
Conference_Location :
Brussels
Print_ISBN :
978-1-4577-1304-0
Electronic_ISBN :
1522-4880
DOI :
10.1109/ICIP.2011.6116406