关键词: 人脸识别, 图像加密, 并行计算, 异构计算

Abstract: To strengthen protecting the privacy of the face recognition system, this thesis introduces an algorithm of regarding image encryption based on changing gray-scale to guard significant recognition objectives. Whats more, to speed up the face recognition and image encryption, the thesis uses OpenCL to realize acceleration with single or multiple devices on heterogeneous platform. The test result shows that the local acceleration of face recognition reaches 1.9x and image encryption acceleration is 6.65x when using the single GPU, while the former statistics is 4.11x and the latter data is close to 6.65x when using dual GPU.


Key words:  , face recognition； image encryption； parallel computing； heterogeneous computing