Abstract:

Stroke extraction is the preprocessing step of the writer identification which is
aimed at distinguishing between different writers. This paper presents an algorithm for the
handwritten character stroke extraction of writer identification. The corresponding
relationship between the discontinuous points of contours and junction regions of strokes is
defined. All junction regions and the junction types are determined by the detection of the
convex and concave discontinuous points in the contours of character images,and the junction
regions are decomposed as their junction types. Strokes are thinned by the method of the
following contour,finally.The experimental results show that the proposed method is
effective and accurate compared to the methods based on thinning and segmenting techniques.
Therefore,the method for extracting the handwritten chinese strokes based on shape
decomposition in this paper features high operational ability and is practical.

Key words: writer identification;stroke extraction;shape decomposition;thinning algorithm