Presentation 2004/12/9
Generative models and model selection for the perception of structure from motion
Nobuhiko ASAKURA, Masao OHMI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We propose a theoretical framework for the perception of 3-D structure from motion that builds on the concepts of a generative model and model selection. This framework claims that the human visual system can make use of generative models of rigid and nonrigid motion for generating predictions of incoming moving images, and select an appropriate model for the recovery of 3-D structure from given moving images based on the compatibility between the predicted and given image data. We present supporting psychophysical evidence that rigid motion displays tend to be perceived as nonrigid when they are incompatible with the predictions of the generative model for rigid motion. We also show that the process of model selection can be viewed as a Bayesian decision process.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Structure from motion / Rigidity assumption / Generative model / Model selection
Paper # HIP2004-64
Date of Issue

Conference Information
Committee HIP
Conference Date 2004/12/9(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Human Information Processing (HIP)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Generative models and model selection for the perception of structure from motion
Sub Title (in English)
Keyword(1) Structure from motion
Keyword(2) Rigidity assumption
Keyword(3) Generative model
Keyword(4) Model selection
1st Author's Name Nobuhiko ASAKURA
1st Author's Affiliation Human Information System laboratory, Kanazawa Institute of Technology()
2nd Author's Name Masao OHMI
2nd Author's Affiliation Human Information System laboratory, Kanazawa Institute of Technology
Date 2004/12/9
Paper # HIP2004-64
Volume (vol) vol.104
Number (no) 525
Page pp.pp.-
#Pages 6
Date of Issue