A trajectory-based computational model for optical flow estimation

Abstract

Abstract-A new computational model for optical flow estimation isproposed. The proposed model utilizes trajectory information present in a multiframe spatio-temporal volume. Optical flow estimation is formulated as an optimization problem. The solution to this optimization problem yields a velocity field corresponding to smoothest and shortest trajectories of constant intensity points within the spatio-temporal volume. The approach is motivated by principles of inertia of morion and least action in physics and vision psychology. An analogy between a trajectory and a "thin wire" is discussed. A simple mechanism for handling trajectory discontinuities is also incorporated. The optimization problem is solved by stochastic relaxation techniques. Some experimental results and performance comparisons with two existing optical flow estimation techniques are presented to demonstrate the effectiveness of the proposed approach.

A Fusion Approach for Multi-Frame Optical Flow Estimation

“…Werlberger et al tackle this problem by using the forward flow from t − 1 and the backward flow from t + 1 in the brightness constancy term for frame t [47]. On the other hand, Chaudhury and Mehrotra pose optical flow estimation as the problem of finding the shortest path with the smallest curvature between pixels in frames of a sequence [13]. Similarly, Volz et al enforce temporal consistency from past and future frames in two ways: accumulating errors across all frames leads to spatial consistency, and trajectories are also encouraged to be smooth [44].…”

Optical Flow Estimation in the Deep Learning Age

“…This property is sometimes also referred to as constant velocity or acceleration assumption. Another way is to parameterize and model the trajectories of motion, which allows to exploit higher-level motion information instead of simply enforcing temporal smoothness on optical flow [9,13,45] in 2D. Recently, there has been initial work on adopting these proven ideas in the context of deep learning to improve the flow accuracy.…”

Development of real-time image sequence analysis for evaluating posture change and respiratory rate of a subject in bed

“…Optical flow is assumed to be constant in the local vicinity, including the spatial and temporal local area (Chaudhury andMehrotra 1995, Zhang et al 1996). Consider the motion of a brightness pattern that is displaced by a distance (δx, δy) in time δt.…”