Shape detection using word like image description

A robust methodology for recognizing distorted and/or partially occluded contours is presented. Two-dimensional models are chosen as the basis of this representation scheme. The models can easily be constructed from a set of prototype objects. The technique is robust enough to tolerate the presence of random noise in the object contours. A semantic approach is used for the recognition task. The scheme fails for scenes with completely occluded objects and for objects of similar shape but of different size