Computational analysis of non-Fourier motion

Area of Science:

• Visual Neuroscience
• Computational Vision
• Perception Psychology

Background:

• Non-Fourier motion is prevalent in visual perception research.
• A robust computational framework for non-Fourier motion is currently lacking.
• Existing models struggle to unify diverse non-Fourier motion phenomena.

Purpose of the Study:

• To develop a unifying theoretical framework for non-Fourier motion perception.
• To characterize non-Fourier motion stimuli using frequency domain properties.
• To explore the computational nature and natural sources of non-Fourier motion.

Main Methods:

• Analysis of non-Fourier motion stimuli in the frequency domain.
• Characterization of stimuli based on oriented power distributions not passing through the origin.
• Examination of spectral power location and orientation using band-pass filter outputs.

Main Results:

• A unifying framework is proposed based on frequency domain power distributions.
• The orientation of spectral power correlates with multiplicative envelope velocity.
• Non-Fourier motion can arise from lighting, translucency, or occlusion effects.
• Spectral power features are extractable from band-pass filter phase and amplitude.

Conclusions:

• The frequency domain characterization provides a unified approach to non-Fourier motion.
• This framework aligns with and extends existing non-Fourier motion models.
• Understanding spectral power offers insights into the computational mechanisms of non-Fourier motion perception.