A real-time, parallel face recognition on task-level pipelined multicomputer
Choong Hwan Lee, Jun Sung Kim, Yeong Ki Jung, Kyu Ho Park · 2002
This paper proposes an efficient face recognition scheme for getting real-time performance. The scheme is characterized by color-based facial region detection and normalized template matching for face identification. For real-time recognition, the authors also implemented a task-level pipelined multicomputer system (RV860-PIPE) on which the proposed face recognition algorithm is implemented. In the experiments with realistic face images, the proposed system shows that the 95% confidence interval for recognition rate ranges from 91%/spl sim/97% typically with 1 frame/sec throughput.