A MODEL-BASED ACTIVE LANDMARKS TRACKING METHOD

Unlike The time-consuming contour tracking method of snakes [5] which requires a considerable number of iterated computations before contour s are successfully tracked down, we present a faster and accurate mode l-based ''landmarks'' tracking method where a single iteration of the dynamic programming is sufficient to obtain a local minimum to an inte gral measure of the elastic and the image energy functionals. The key lies in choosing a relatively small number of salient ''landmarks'', o r features of objects, rather than their contours as a target of track ing within the image structure. The landmarks comprising singular poin ts along the model contours are tracked down within the image structur e all inside restricted search areas of 41 x 41 pixels whose respectiv e locations in image structure are dictated by their locations in the model. A Manhattan distance and a template corner detection function o f Singh and Shneier [7] are used as elastic energy and image energy re spectively in the algorithm. A first approximation to the image contou r is obtained in our method by applying the thin-plate spline transfor mation of Bookstein [2] using these landmarks as fixed points of the t ransformation which is capable of preserving a global shape informatio n of the model including the relative configuration of landmarks and c onsequently surrounding contours of the model in the image structure. The actual image contours are further tracked down by applying an acti ve edge tracker using now simplified line search segments so that indi vidual differences persisting between the mapped model contour are sub stantially eliminated. We have applied our method tentatively to portr aits of a class album to demonstrate the effectiveness of the method. Our experiments convincingly show that using only about 11 feature poi nts our method provides not only a much improved computational complex ity requiring only 0.94 sec. in CPU time by SGI's indigo2 but also mor e accurate shape representations than those obtained by the snakes met hods. The method is powerful in a problem domain where the model-based approach is applicable, possibly allowing real time processing becaus e a most time consuming algorithm of corner template evaluation can be easily implemented by parallel processing firmware.