作者: Do Joon Jung , Kyung Su Kwon , Hang Joon Kim
DOI: 10.1007/978-3-540-73110-8_71
关键词:
摘要: In this paper, we propose an approach toward body parts representation, localization, and human pose estimation from image. the image, a background are represented by mixture of Gaussians, configuration is modeled Bayesian network. model, state nodes represent parameters each part, arcs spatial constraints. The Gaussian distribution used to model prior for as parametric model. We estimate through optimization using likelihood objective functions. performance proposed illustrated on various single images, improves quality.