Abstract
In this paper, we propose a method to utilize the color texture of an object for estimating the 3D posture of the human body from monocular photography image. In the proposed method, first, we map the object texture to a 3D human body model. Next, we assign postures in the motion capture database on the human body model and produce posture images by rendering the model from eight horizontally-surrounding directions. Finally, we compare the photography image with these posture images, and estimate the posture of the human body in the photography image. We compare the posture estimation results obtained by using RGB value, HOG feature and color histogram, and discuss the effectiveness of the proposed method.
