Paper Info
Title
Human distribution estimation using shape projection model based on multiple-viewpoint observations
Abstract
This paper describes a method for estimating human distributions (quantities and locations) based on multiple-viewpoint image sequences. In the field of human image analysis, inter-human occlusion is a significant problem: when a scene includes a large number of occlusions, tracking of individual persons becomes difficult. Therefore, updating a tracking-based model is not enough to estimate the distribution in complex scenes. In our method, the number of persons and their locations are directly estimated from a set of input images based on the fitting of a projected shape model. The model’s complexity (number of persons) is determined based on the MDL (minimum description length) criterion. In addition, the image areas occluded by static objects are also detected and automatically excluded from the human distribution computations. We confirmed the feasibility of the proposed method through experiments using both synthesized and real images. Results show the effectiveness of our method.
Year
DOI
Venue
2006
10.1007/11612032_80
ACCV (1)
Keywords
Field
DocType
image area,human image analysis,human distribution computation,multiple-viewpoint observation,real image,human distribution estimation,input image,large number,human distribution,shape projection model,projected shape model,multiple-viewpoint image sequence,distributed computing,image analysis,minimum description length
Computer vision,Model matching,Pattern recognition,Occultation,Computer science,Minimum description length,Image processing,Artificial intelligence,Real image,Image sequence,Computation,Computational complexity theory
Conference
Volume
ISSN
ISBN
3851
0302-9743
3-540-31219-6
Citations 
PageRank 
References 
0
0.34
12
Authors
4
Name
Order
Citations
PageRank
akira utsumi129039.43
Hirotake Yamazoe212624.07
Kenichi Hosaka311015.66
Seiji Igi48314.01