Adaptive Step-Size Fast Interested Boundary Tracking

Traditional boundary tracking finds large number of redundant edge points and consumes more time. To resolve these disadvantages, an adaptive step-size boundary tracking algorithm is proposed. The new algorithm estimates possible location of next node and search edge point within extended neighborhood of the estimated node. Moreover, the new method adjusts searching step size according to current boundary curvature to make sure the detected points distributed more scientifically. Experiments were performed and the results shown that the new algorithm is better than alterable window reptile method in detected nodes number and processing speed. Compare with alterable window reptile tracking method, nodes distribution of our algorithm is accord with the characteristic of detected boundary and the number of gotten points and searching time is reduced by 95%-96% and 45%-84%. These results suggest that the new algorithm can be used to detect interested contour of large-scale digital image.