Probabilistic Motion Estimation Based on Temporal Coherence

TL;DR

This paper presents a Bayesian-based theory for temporal motion estimation that groups visual data over time, predicting motion flows and modeling cortical-like neural networks, aligning with psychophysical findings.

Contribution

It introduces a novel Bayesian temporal grouping framework for motion estimation, extending data association methods with a neural network implementation.

Findings

Qualitatively explains psychophysical experiments on motion occlusion.

Demonstrates the neural network model predicts motion outliers.

Shows effectiveness of temporal grouping in motion prediction.

Abstract

We develop a theory for the temporal integration of visual motion motivated by psychophysical experiments. The theory proposes that input data are temporally grouped and used to predict and estimate the motion flows in the image sequence. This temporal grouping can be considered a generalization of the data association techniques used by engineers to study motion sequences. Our temporal-grouping theory is expressed in terms of the Bayesian generalization of standard Kalman filtering. To implement the theory we derive a parallel network which shares some properties of cortical networks. Computer simulations of this network demonstrate that our theory qualitatively accounts for psychophysical experiments on motion occlusion and motion outliers.