In this paper, we are dealing with the analysis of dynamic scenes filmedudby a féxed camera . The purpose of thés analysis is a quick locating of theudmoving abjects, without prior knowledge of the scene and of the nature ofudthe movement. The existing methods, which best respond to these criteria,udcompare successive images in order to exploit edge position changes .udHowever, each of these methods misinterprets some of the static edgesudwhich are flot visible in all the compared images. As a matter offact, theudinformation provided by three consecutive images is sufficient to tell theseudstatic edges from the real moving edges .We are proposing, herein, an original operator which locates the movingudedges in the carrent image by comparing the gradient of thés image to theudgradients of the previous image and of the next one . Our method bothudimproves the elimination of the static edges and makes it possible toudextract a higher number of pertinent edges .
展开▼
