Real-time Face Detection Algorithm Using GPU
Processing of human faces plays an important part in various domains such as biometrics, video surveillance, information security, interactive video games and human computer interaction; most of them require real-time and interactive processing. In this paper, we present a new GPU implementation of human face detection algorithm based on OpenCL. Compared with CPU version, it extraordinarily improves the performance and detection speed. Besides, it also implements high acceleration ratio. With the difference of other GPU-based algorithms, we have done many other new parallelization works independently, such as parallel processing of Haar wavelet, cascade classifier and different sizes or positions of detected sub-windows. Experimental data show that our proposed method achieves better effect. The performance of GPU version is enhanced, detection speed is accelerated up to 10 to 42 times compared with CPU-based implementation, also real-time level constraints is met as well.
face detection AdaBoost parallel GPU real-time
Zhongyuan FENG Jinyuan JIA Feipeng ZHAO
School of Software Engineering Tongji University Shanghai, China
国际会议
重庆
英文
1-6
2011-12-01（万方平台首次上网日期，不代表论文的发表时间）