3-D MOTION ESTIMATION FROM OPTICAL-FLOW WITH LOW COMPUTATIONAL COST AND SMALL VARIANCE

In this paper, we study three-dimensional motion estimation using opti cal flow. We construct a weighted quotient-form objective function tha t provides an unbiased estimator. Using this objective function with a certain projection operator as a weight drastically reduces the compu tational cost for estimation compared with using the maximum likelihoo d estimator. 70 reduce the variance of the estimator, we examine the w eight, and we show by theoretical evaluations and simulations that, wi th an appropriate projection function, and when the noise variance is not too small, this objective function provides an estimator whose var iance is smaller than that of the maximum likelihood estimator. The us e of this projection is based on the knowledge that the depth function has a positive value (i.e., the object is in front of the camera) and that it is generally smooth.