Human body pose estimation using silhouette shape analysis

Abstract

We describe a system for human body pose estimation from multiple views that is fast and completely automatic. The algorithm works in the presence of multiple people by decoupling the problems of pose estimation of different people.The pose is estimated based on a likelihood function that integrates information from multiple views and thus obtains a globally optimal solution. Other characteristics that make our method more general than previous work include: (1) no manual initialization, (2) no specification of the dimensions of the 3D structure, (3) no reliance on some learned poses or patterns of activity, and (4) insensitivity to edges and clutter in the background and within the foreground. The algorithm has applications in surveillance and promising results have been obtained.

Statistical Motion Information Extraction and Representation for Semantic Video Analysis

“…It must be noted that for this particular task a series of dedicated approaches, exhibiting high recognition rates, have already been presented in the literature. The proposed method, although it does not exploit specific facts and characteristics that are only present in this domain (like human body silhouette extraction [56], body pose estimation [57], etc. ), which can significantly facilitate the recognition procedure, nevertheless presents satisfactory results.…”

Constraint Integration for Efficient Multiview Pose Estimation with Self-Occlusions

“…al [43] use a formulation similar to the above [47] for handling self-occlusions, but with a different likelihood model, and apply it to 2D pose estimation. April 11, 2007 DRAFT The pose estimation problem can be simplified significantly by assuming that the person can be segmented from the image, say using background subtraction [4], [23], [29]. While this reduces the search space significantly, it does not handle the problem of self-occlusion or people occluding one another.…”

Markov-based Silhouette Extraction for Three--Dimensional Body Tracking in Presence of Cluttered Background

“…Using edges obtained by gradient based methods for example as in [5,7,9,6] is appealing due to its simplicity and speed, but its application is restricted to the cases where the contrast is sufficient. Furthermore, these methods tend to fail in the presence of highly textured objects and clutter, which produce too many irrelevant edges.…”

“…All methods for inferring 3-D pose of articulated models such as whole bodies [1,7,5,6,9,8] or hands [12,2,14] using occluding contours depend on reliably extracting the outlines of the subject. For example, Kakadiaris and Metaxas [8] establish direct correspondences between points on the occluding contours to points on the surface of the model using projective geometry.…”

“…A number of promising silhouette-based approaches to body tracking and human pose estimation have been proposed recently [1,5,6,9,13]. However most of them rely on the fact that silhouettes can be extracted using relatively simple algorithms such as background subtraction [4] or standard edge-and gradient-based techniques [7,2,1].…”