基于多分辨率方法的主动轮廓线跟踪算法

TRACKING ALGORITHM FOR ACTIVE CONTOURS BASED ON MULTIRESOLUTION

由Kass等于1987年首次提出的主动轮廓线模型，在数字图像分析和计算机视觉领域得到了越来越广泛的应用．基于单分辨率的主动轮廓线跟踪算法存在一个共同的缺点．即：要求初始轮廓线离目标的真实轮廓线很近，这样的算法用于跟踪将只能跟踪缓慢运动的目标，从而限制了主动轮廓线跟踪算法的应用范围．本文提出了一种基于多分辨率的主动轮廓线跟踪算法，它不但继承了基于单分辨率的主动轮廓线算法的优点，而且降低了对初始轮廓线的要求．此外，本文推导出不同分辨率情况下主动轮廓线算法与其允许的目标运动速度的关系公式，并用这些公式定量地计算出单分辨率和多分辨率情况下算法允许目标运动的最大速度，证明本文方法可以跟踪运动较快的目标．

Active contour models, introduced by Kass M et al in 1987, have been used more and more widely in applications of image analysis and computer vision.However, active contour models based on monoresolution have a common weak-ness, i. e., they require that the initial contour is very close to the true contour of the target, such tracking algorithms can only track targets in slow speed which lim-its the application area of these tracking algorithms. In this paper, a new tracking algorithm for active contours based on multiresolution is presented, which not only retains the advantages of the existing algorithms based on monoresolution, but also cuts down the strict requirement to initial contours. Moreover, the formulas of re-lations between the tracking algorithms and the velocity of the target allowed by the algorithms in different resolution are derived. According to these formulas, the maximum allowed velocity of the target may be estimated under both monoresolu-tion and multiresolution cases. The analyzed result proves that the proposed method can track the target in faster speed.