Discrimination of speed distributions: sensitivity to statistical properties

Area of Science:

• Perception and Cognition
• Computational Neuroscience
• Computer Vision

Background:

• Human observers process complex visual information, including motion cues.
• Understanding the statistical properties of velocity distributions is key to visual perception.
• Previous research has explored visual sensitivity to statistical properties.

Purpose of the Study:

• To investigate human ability to discern statistical differences in velocity distributions.
• To determine which statistical moments (mean, variance, skewness, kurtosis) are detectable.
• To assess the implications for computer vision tasks like image segmentation and heading detection.

Main Methods:

• Two experiments employed a four-alternative forced-choice methodology with simultaneous velocity distributions.
• Experiment 1 manipulated one statistical moment (mean, variance, skewness, or kurtosis) while holding others constant.
• Experiment 2 focused on variance, skewness, and kurtosis, providing trial-by-trial feedback.

Main Results:

• Human observers reliably detected differences in the mean of velocity distributions.
• Human observers reliably detected differences in the variance of velocity distributions.
• Sensitivity to skewness and kurtosis differences was less pronounced or not explicitly detailed as reliably detected.

Conclusions:

• Human visual system is adept at processing statistical information within velocity distributions.
• Detection of mean and variance differences has direct applications in computer vision.
• Findings inform algorithms for image segmentation and heading detection from optic flow.