Acceleration algorithm for CUDA-based face detection

For the traditional CPU facial detection program is difficult to meet the requirements of real-time detection over high definition images, we propose a fast face detection acceleration algorithm based on the Viola-Jones cascade classifier in the Compute Unified Device Architecture (CUDA) platform. We implemented and improved novel parallel methodologies of image integral calculation, scan window processing and the amplification and correction of classifiers. The unbalanced workload problem of irregular tree structure applications could be effectively solved by the window re-mapping technology using in the scan window processing. The experimental results show that, the CUDA program running on an Nvidia GTX 570 graphics card could respectively achieve 17.04, 3.22 times speedup compared with the CPU and Open Source Computer Vision (OpenCV) programs running on an Intel core i7 processor for the 1080p image, while maintaining a similar detection quality.