基于GPU的Gabor人脸图像特征提取算法的研究

摘要

In this paper an algorithm for Gabor wavelet feature extraction based on GPU parallel computing is proposed to improve the traditional Gabor wavelet characteristics in real-time of face feature extraction. The basic ideal of this algorithm is to operate convolution of Gabor wavelet and face image on GPU in parallel by CUDA program model and multi-thread mapping. The experiment results show that the proposed algorithm improves the speed of calculation at 12 times compared with Gabor-CPU Programming, and providing effective technical support for human face features extracting in real time system.%论文针对传统Gabor小波计算在人脸图像特征提取中实时性较弱的问题,提出了一种基于GPU并行计算的Gabor小波特征提取方法.所提算法将Gabor小波与人脸图像的卷积运算在GPU(Graphic Processing Unit图形处理器)上并行实现,并采用CUDA (Com pute Unified Device Architecture)编程模型,利用多线程并行映射实现.与传统的Gabor小波人脸特征提取算法对比实验表明,文中方法的计算速度在CPU上速度提高了近12倍,为人脸特征实时提取提供了有效的技术保障.