Abstract
This study focuses on the problem of human face pose estimation based on single image. Traditional methods for 2D-3D feature based pose estimation problem require two inputs, and they cannot work well due to lack of correspondences of input images. We transfer the problem into an optimization problem via six-point template, and solve the problem by time-varying acceleration coefficients particle swarm optimization (TVAC-PSO). Experiments on 40 different poses demonstrate that the TVAC-PSO is superior to either GA or PSO in terms of accuracy.