Writer Identification Based on Local Contour Distribution Feature

Abstract

A method based on local contour distribution features is proposed for writer identification in this paper. In preprocessing, contours are abstracted form images by an improved Bernson algorithm. Then the Local Contour Distribution Feature (LCDF) is extracted from the fragments which are parts of the contour in sliding windows. In order to reduce the impact of stroke weight, the fragments which do not directly connect the center point are ignored in the feature abstraction procedure. The edge point distributions of the fragments are counted and normalized into LCDFs. At last, the weighted Manhattan distance is used as similarity measurement. The experiments on our database and ICDAR 2011 writer identification database show that the performance of the proposed method reach or exceed those of existing state-of-art methods.