No abstract

No abstract

This paper presents an analysis study of handwritten Kanji characters on a digitizing tablet. To solve a writer identification and verification problem, we have tried to extract significant individual characteristics of a certain set of Kanji characters, which might be defined as feature parameters derived from the knowledge of document examiners. A set of feature parameters in terms of character shape and writing behavior is evaluated by the statistical test. Evaluation results have shown that some feature parameters are good enough to identify and verify a writer himself.

In automatic computer verification of seal imprints accurate positioning of an examined seal imprint over a registered imprint is of great importance. This paper proposes a method which positions an examined seal imprint in relation to a parallel shift and rotation by using the centroid of its frame figure and the pixels on a circle of a certain radius from the centroid. Since this method achieves positioning by using local features of the examined seal imprint, rapid positioning is possible. In the presence of small defects in the imprint frame this method has the advantage that the centroid of the seal imprint is accurately obtained and is not affected by distortion of the imprint frame. In a positioning experiment based on this method 240 actual seal imprints were used. All imprints were positioned within a ±0.1‐mm parallel shift error and more than 99% of the seal imprints were positioned within ±1.76° rotation angle errors. A verification experiment was conducted using a seal imprint verification system which incorporated the principles of this method. As a result, the verification error which was caused by the positioning error was found to be 0.83%. The experimental results show that the proposed positioning method is efficient for automatic verification of seal imprints.