|
ABSTRACT In this paper, we present an object contour tracking approach using graph cuts based active contours (GCBAC). Our GCBAC based tracking approach does not need any a priori global shape model, which makes it useful for tracking objects with deformable shapes and varying appearances. Central to the tracking approach is the use of our GCBAC based segmentation algortihm which extracts object contours by starting with an initial estimate and iteratively converging to the optimal contour within its own neighborhood. In the first frame of a video sequence, GCBAC algortihm iteratively converges to an optimal object boundary from a given initial contour. In each frame thereafter, the tracking algorithm occurs in two steps. In the first step, GCBAC algorithm is applied to the difference between this frame and the previous frame, using the resulting contour from the previous frame as initialization. This yields an estimate of the moving area. In the second step, GCBAC algorithm is applied directly to current frame, using the resulting contour from step 1 as initialization. This yields an optimal object contour. We present experimental results on several real world video sequences. DISCUSSION
|
