Tracking Object Contour Using A Novel B-Snake Algorithm
QIU Shu-bo,WANG Hua-xiang,LIANG Zhi-wei
DOI: https://doi.org/10.3969/j.issn.1006-8961.2005.05.007
2005-01-01
Journal of Image and Graphics
Abstract:In some cases, the object contour under observation may be adjacent to some other objects in the background when using Snake algorithm to track moving object. If not be constrained, some control points might be drawn to the background boundary, which result in an error. On the other hand, if the object is moved or rotated, contour control points may result in crossing, which leads to failure of contour tracking. Thus, a novel and efficient algorithm is proposed in this paper to solve those problems mentioned above. The proposed algorithm can be used to track 2D contour for both static and moving objects. The contour is initialized by connecting control points using cubic B-Spline curve.Then control points are drawn to the contour boundary using minimum energy function. The B-Snake tracking algorithm adopts the sine value of the angle between the vector representing the next point and the vector representing the control center point to prevent control points from crossing. And the position of the control center point can be estimated efficiently by α-β-γ filtering. Therefore, moving object contour can be tracked precisely. Experimental results of applying the proposed algorithm in static or moving human head contour tracking show that the proposed algorithm yields good effect and saves time.