Related Experiment Videos
The perceptual buildup of three-dimensional structure from motion
E C Hildreth1, N M Grzywacz, E H Adelson
1Massachusetts Institute of Technology, Cambridge.
Perception & Psychophysics
|July 1, 1990
Summary
Human perception accurately reconstructs 3-D structure from motion, improving over time. This visual processing is robust to noise and influenced by prior structural models.
Area of Science:
- Cognitive Psychology
- Computational Neuroscience
- Computer Vision
Background:
- Investigating the human visual system's ability to perceive three-dimensional (3-D) structure from dynamic 2-D image sequences.
- Exploring the computational principles underlying 3-D structure perception from motion, referencing Ullman's incremental rigidity scheme.
Purpose of the Study:
- To psychophysically measure the accuracy of 3-D structure perception derived from relative motion.
- To compare human performance with predictions from the incremental rigidity computational model.
Main Methods:
- Conducted psychophysical experiments measuring the accuracy of perceived 3-D structure from changing 2-D images.
- Introduced noise into image positions to assess robustness.
- Utilized computer simulations to compare experimental results with Ullman's incremental rigidity model.
Main Results:
- Demonstrated that the human visual system can accurately derive relative depth information from moving points, even with noisy image data.
- Showed that 3-D model accuracy improves over time, eventually reaching a stable plateau.
- Found that current 3-D structure perception is influenced by previously perceived models.
Conclusions:
- The human visual system effectively constructs accurate 3-D structural models from relative motion.
- Perceptual accuracy is time-dependent and robust to image noise, aligning with incremental processing principles.
- Prior perceptual experiences modulate the ongoing construction of 3-D structure from motion.