Accelerating Active Shape Model using GPU for facial extraction in video

In this paper, we present a novel parallel implementation of Active Shape Model (ASM) on GPU for massive facial feature extractions in video or image sequence. With the Compute Unified Device Architecture (CUDA)-enabled GPU, the acceleration is significant and reported a 48 times performance boost comparing to a CPU implementation. We adopt the hardware built-in bilinear interpolation of texture to shorten the time for a large number of image scale transform operations. Then, a GPU-based parallel mahalanobis distance calculation is introduced in the searching process, and this enables most of the computations to be performed simultaneously. As a result, we can achieve real-time performance in our video-driven 3D facial animation system.