Human pose estimation based on constraint pictorial structure model

Aiming at the problem of estimating human pose about the issue of human motion capture from video images, this paper proposes a new method for human pose estimation based on constraint pictorial structure model. In order to reduce the searching space of estimation algorithm, the location and the size of the people are firstly detected through a simple tracking algorithm, Then, a Grab-Cut algorithm is used to extract the foreground information, which can be used as the input of the estimation stage. And we extend the general pictorial structure model, finally we apply a prior segmentation in the estimation stage. As a result, the searching space is reduced a lot, for we don´t need to search the whole image. The simulation results demonstrate the proposed method is effective.