Human Upper Body Pose Estimation in Static Images

Estimating human pose in static images is challenging due to the high dimensional state space, presence of image clutter and ambiguities of image observations. We present an MCMC framework for estimating 3D human upper body pose. A generative model, comprising of the human articulated structure, shape and clothing models, is used to formulate likelihood measures for evaluating solution candidates. We adopt a data-driven proposal mechanism for searching the solution space efficiently. We introduce the use of proposal maps, which is an efficient way of implementing inference proposals derived from multiple types of image cues. Qualitative and quantitative results show that the technique is effective in estimating 3D body pose over a variety of images. 1 Estimating Pose in Static Image This paper proposes a technique for estimating human upper body pose in static images. Specifically, we want to estimate the 3D body configuration defined by a set of parameters that represent the global orientation of the body and body joint angles. We are focusing on middle resolution images, where a person’s upper body length is about 100 pixels or more. Images of people in meetings or other indoor environment are usually of this resolution. We are currently only concerned with estimating the upper body pose, which is relevant for indoor scene. In this situation the lower body is often occluded and the upper body conveys most of a person’s gestures. We do not make any restrictive assumptions about the background and the human shape and clothing, except for not wearing any head wear nor gloves.