A unified model of short-range and long-range motion perception

Xuming He (NICTA)

COMPUTER VISION AND ROBOTICS SERIES

DATE: 2011-03-10
TIME: 16:00:00 - 17:00:00
LOCATION: RSISE Seminar Room, ground floor, building 115, cnr. North and Daley Roads, ANU
CONTACT: JavaScript must be enabled to display this email address.

ABSTRACT:
The human vision system is able to effortlessly perceive both short-range and long-range motion patterns in complex dynamic scenes. It has been proposed that different mechanisms are involved in processing these two types of motion. We proposed a hierarchical model as a unified framework for modeling short-range and long-range motion perception. Our model consists of two key components: a data likelihood that proposes multiple motion hypotheses using nonlinear matching, and a hierarchical prior that imposes slowness and spatial smoothness constraints on motion field at multiple scales. We tested our model on two types of stimuli, random-dot kinematogram and multiple-aperture stimulus, both commonly used in human vision research. The hierarchical model adequately accounts for human performance in the psychophysical experiments.
BIO:
Xuming He is currently a Researcher at NICTA Canberra Research Laboratory. His research interests mainly focus on computer vision and machine learning. Recent work has been on scene segmentation with conditional random fields and latent topic models; Motion estimation and occlusion detection with hierarchical Markov network.

He studied Electronics Engineering (BSc & MSc) at Shanghai Jiaotong University, China; and received his PhD of computer science from University of Toronto, Canada. Before he joined NICTA in 2010, he was a postdoc researcher at Center for Image and Vision Science in UCLA, USA.