| One of the central problems in pattern recognition and computer vision is the representation, description, and matching of shapes. When a pattern to be matched is allowed to be in variable locations and orientations, the registration problem occurs. The registration problem involves determining a transformation that maps arbitrary point p in the first image coordinate space to corresponding point q in the second image coordinate space.; For the purpose of entity identification, seals, also called chops, signets, or stamps, are extensively used in Korea, China, Japan, and other oriental countries. The entity may be a person, a social group, or a department of the government. Seals can be different in size and shape. In addition, seal imprints can be set in variable locations and orientations, and could be in different scales when variable focusing is allowed. Therefore, without the pre-knowledge regarding the location, orientation, and scaling information of the seal imprints, there certainly exists the registration problem in the seal imprint verification process. Depending on the stamping conditions, seal imprints could also have a lot of variations even they are produced from a single seal.; This research project investigates the use of Medial Axis Transformation (MAT) to solve the Seal Imprint Verification problem. Specifically, this research project is interested in verifying seal imprints in variable locations, orientations, and scales. The MAT is a technique that can reduce a planar object to a skeleton. From the skeleton and the Euclidean elevation map, critical features can be derived. Using a relaxation technique incorporated with the features derived from the skeleton, the registration problem can be resolved. Finally, by introducing the piecewise parametric cubic curve for the representation of skeleton segments, seal imprints can be verified with greater flexibility and better success. |
