Abstract

The effectiveness of the use of stochastic gradient estimation to detect motion in a sequence of images is investigated. Pixel-by-pixel reverse estimation of the deformation field is used. The representation of the shift vector is considered both by the projections on the basic axes and by the polar parameters. Two approaches to estimate the parameters of the deformation field are proposed and analyzed. In the first approach, the stochastic gradient procedure sequentially processes all rows of an image to find estimates of shifts for all points of the reference image. The joint processing of the results allows compensating the inertia of the stochastic estimation. In the second approach, to improve the accuracy of estimation, the correlation of image rows is taken into account. As a criterion for the formation of the resulting estimate, the minimum of gradient estimation and correlation maximum were investigated. The computational complexity of the proposed algorithms is investigated. The algorithms are compared with the MVFAST algorithm. Examples of experimental results on the formation of the deformation field, the selection of a moving object area, and the finding of the movement and trajectory parameters of a moving object in a video sequence are given.

Keywords: image processing, video sequence, motion, shift vector, detection, stochastic estimation, stochastic gradient, trajectory, reverse estimation, correlation
