Human body pose estimation from still images and video frames

Authors:
Amar A. El-Sallam;Ajmal S. Mian
Affiliations:
School of Computer Science and Software Engineering, University of Western Australia, WA;School of Computer Science and Software Engineering, University of Western Australia, WA
Venue:
ICIAR'10 Proceedings of the 7th international conference on Image Analysis and Recognition - Volume Part I
Year:
2010

Citing 10
Cited 0

A Survey of Glove-based Input

IEEE Computer Graphics and Applications
Towards model-based recognition of human movements in image sequences

CVGIP: Image Understanding
Probabilistic Methods for Finding People

International Journal of Computer Vision
Computer Vision

Computer Vision
Beyond Trees: Common-Factor Models for 2D Human Pose Recovery

ICCV '05 Proceedings of the Tenth IEEE International Conference on Computer Vision (ICCV'05) Volume 1 - Volume 01
Recovering 3D Human Pose from Monocular Images

IEEE Transactions on Pattern Analysis and Machine Intelligence
Measure Locally, Reason Globally: Occlusion-sensitive Articulated Pose Estimation

CVPR '06 Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition - Volume 2
3D Skeleton-Based Body Pose Recovery

3DPVT '06 Proceedings of the Third International Symposium on 3D Data Processing, Visualization, and Transmission (3DPVT'06)
Vision-based hand pose estimation: A review

Computer Vision and Image Understanding
ViHASi: virtual human action silhouette data for the performance evaluation of silhouette-based action recognition methods

VNBA '08 Proceedings of the 1st ACM workshop on Vision networks for behavior analysis

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper presents a marker-less approach for human body pose estimation. It employs skeletons extracted from 2D binary silhouettes of videos and uses a classification method to partition the resultant skeletons into five regions namely, the spine and four limbs. The classification method also identifies the neck, the head and the shoulders. Using the center of mass principles, a model is fitted to the body parts. The spine is modeled with a 2nd order curve while each limb is modeled by two intersected lines. Finally, the model parameters represented by a reference point and two angles belonging to the lines are estimated and the pose is reconstructed. The proposed approach can estimate body poses from single images as well as multiple frames and is considerably robust to occlusions. Unlike existing methods, our approach is computationally efficient and can track human motion while correcting for pose errors using multiple frames. The proposed approach was tested on real videos from MuHAVi and MAS databases and gave promising results.