A comparison of motion integration for optic flow components

Alan LEE, Hongjing LU

Research output: Journal PublicationsConference Abstractpeer-review


In order to perceive optic flow from moving objects and/or scenes, the visual system integrates local motion signals over space and time to form a globally coherent motion percept. Our previous study revealed greater sensitivity in perceiving circular/radial motion than for translational motion (Lee et al., 2008), suggesting the existence of specialized motion integration mechanisms for analyzing complex “optic flow” motion. The current study examined the properties of integration mechanisms tuned to different motion patterns, comparing their spatial and temporal extent.

Stimulus consisted of 728 drifting Gabor elements, each with randomly assigned orientation. For signal Gabor elements, drifting velocities were manipulated to generate translational, circular or radial motion. For noise Gabor elements, velocities were random. Sensitivity was measured by proportion of signal elements yielding a performance level of 75% correct in a discrimination task.

In Experiment 1, we found greater sensitivity for circular motion than for translation over a range of speeds, 0.8, 1.6 and 2.4 deg/s. The difference in sensitivities between translational and radial motions diminished with increase in speed. Furthermore, for circular motion, but not translational and radial motion, sensitivity remained fairly constant across different speeds, suggesting that an integration mechanism specialized for circular motion may be tuned for a broader range of speeds. In Experiment 2, we found near invariance of human sensitivity with element density (different numbers of Gabor elements within a 12°circular visual field) for all three motion patterns, suggesting linear pooling of local motion guided by specialized integration mechanisms for translation, circular and radial motion. In Experiment 3, we found that human sensitivity increased linearly with stimulus duration, up to about 80∼150ms, for all the three motion patterns. This result reflects a temporal integration limit in the early stage of local-motion analysis with a time constant of about 100ms.
Original languageEnglish
Pages (from-to)633
JournalJournal of Vision
Issue number8
Publication statusPublished - 24 Mar 2010
Externally publishedYes


Dive into the research topics of 'A comparison of motion integration for optic flow components'. Together they form a unique fingerprint.

Cite this