Recognition of writers by handwriting images

The objective of this paper is the automatic identification of individual persons by means of their handwriting images. The image, digitized by a computer-controlled TV-camera, is analyzed in two different ways. In terms of the writing process, it is first interpreted as a set of lines, which have to be reconstructed by methods of scene analysis.

Secondly, a special non-linear transformation is applied to the image, which provides features representing global statistical information about the image. In a classification experiment involving 20 writers (40 samples of each), 797 out of 800 samples were identified correctly.