TRADEMARK SHAPES DESCRIPTION BY STRING-MATCHING TECHNIQUES

This work considers the possibility of measuring the shape similarity distances of two-dimensional objects. The method proposed is articulated in two stages: chain-coding the objects' contours, and applying string distances to the coded contours. The distance values obtained by such a procedure are then associated to the general geometrical characteristics of the object. Comparison of the distance values of two objects gives an indication of their similarity. The implementation of a shape similarity distance based on this concept is considered in depth. Its theoretical and practical implications are examined. The effectiveness of the method is tested with trademarks, a class of images of considerable practical interest. The results are encouraging and point to the feasibility of the task undertaken by using the proposed approach.