Writer Identification Using Differential Chain Codes and Grid Features
This paper proposes an efficient method for writer identification in handwritten documents. The method operates at two stages: rough matching and fine matching. At the first stage, the histograms of the first and second order differential chain codes are
- PDF / 692,119 Bytes
- 10 Pages / 439.37 x 666.142 pts Page_size
- 46 Downloads / 182 Views
Abstract This paper proposes an efficient method for writer identification in handwritten documents. The method operates at two stages: rough matching and fine matching. At the first stage, the histograms of the first and second order differential chain codes are extracted and used in rough matching. At the second stage, the histograms of the positions of connected edge pixel pairs in a floating grid are recorded and applied to the fine matching. At each stage, two writings are compared by computing the distances between their respective histograms. The experiments are implemented on two common data sets: HIT-MW and IAM, which contain samples of Chinese and English handwritten texts. The experimental results show that the proposed methods achieve promising results on writer identification. Keywords Writer identification Fine matching
Global level Local level Rough matching
1 Introduction The identification of a person on the basis of scanned images of handwriting is a useful biometric modality with application in forensic and historic document analysis and constitutes an exemplary study area within the research field of
Z. Huang D. Wang (&) The Third Research Institute of the Ministry of Public Security, Shanghai, China e-mail: [email protected] Y. Lu Department of Computer Science and Technology, East China Normal University, Shanghai, China
Z. Wen and T. Li (eds.), Foundations of Intelligent Systems, Advances in Intelligent Systems and Computing 277, DOI: 10.1007/978-3-642-54924-3_61, Springer-Verlag Berlin Heidelberg 2014
647
648
Z. Huang et al.
behavioral biometrics [1]. Writer identification methods can be categorized into two types: text-dependent methods and text-independent methods. In text-dependent methods, the writers have to write the same fixed text to be used to establish the identity of writer, but in text-independent methods any text with different characters to perform identification. These methods can be classified into online (also called dynamic), where the information on the writing order and dynamics of the writing process is available, and offline (also called static), where only a scanned image of handwriting is available, and thus much dynamic information of writing process is lost. Research in writer identification has received and renewed interest in the recent years. A wide variety of features have been proposed to distinguish the writing of an individual from another. Said et al. [2] presented a global approach and treated each writer’s handwriting as a different texture using multichannel Gabor filtering and gray-scale co-occurrence matrices techniques. Zois et al. [3] employed a feature vector by morphologically processing horizontal projection profiles of the words. Bulacu et al. [4] evaluated the performance of edge-based directional probability distributions as features in writer identification. Schlapbach et al. [5] proposed an HMM-based approach for writer identification and verification, with a separate system for each writer. The recognizer output log-likel
Data Loading...
