A robust parametric active contour based on fourier descriptors

This paper presents an optimal parametric active contour method based on Fourier descriptors to detect and track a contour from a sequence of images in real-time. The advantage of this approach is that not only convex but also concave contours can be considered. We also propose a method to initialize the active contour using only three user´s clicks in the first image of the sequence and to automatically re-initialize it when possible topological changes are detected during the tracking. Moreover, the algorithm is implemented on GPU to ensure the real-time tracking performance.