Motion Object’s Contour Tracing Based On Optical Flow And Level Set

Based on the analysis of the characteristics of moving image sequence, the motionobject’s contour tracking method based on optical flow theory and level set technique wasput forward in this paper. The method combined the moving target’s optical flowinformation with C-V image segmentation model, and optical flow velocity was treated asthe curve evolution force. The C-V model with optical flow force (abbre. CV model) coulddetect moving targets accurately. To offset the shortage of CV model, the followingimprovements was presented in the research process:First, due to the narrow range definition of Dirac functionδ(φ)in the CV model, therewould be some restrain when it was used to detect edge of the object with holes. Animproved CV model that Dirac function was replaced withδ(φ)was proposed in order tosolve this problem. Thus, the restrain that the edge far away from the active contours in CVmodel was eliminated, and the CV model could fast the iteration greatly when detecting thecontour of moving target.Second, reinitialization is a necessary step to make the level set function as signeddistance function. In this paper the deviation of level set function from the signed distancefunction is introduced into the CV model to keep the level set function as a signed distancefunction during the curve evolution, which made the CV model without re-initialization.Moreover, the level set function can be flexibly initialized with the binary image rather thana signed distance function in practice.Numerical results showed that the first improved method improves the speed of thecurve evolution, and it has a better global optimization characteristics; the second one is lesstime-consuming for complicated initial contour.