Scholarlink

<p>Writer identification problem has become prevalent topic in the field of handwriting biometrics. This paper proposes a new approach for writer identification of handwritten documents. The scanned handwritten images are represented by writing descriptors. In order to explain the intra- and inter-writer variability by computing the similarity measurements, two statistical texture descriptors for characterizing the writers’ handwriting styles has been analyzed. Therefore, the joint probability of LCP-IWSL and LBLP on various pixels is calculated. After that, a combination of these descriptors is performed using the sum rule fusion. Identification is carried out using k-Nearest Neighbors and the Chi-Square distance with the simplest kind of cross validation: Holdout. The proposed scheme achieves interesting performances, according to the experimental results on nine well-known handwriting databases, including two Arabic (IFN/ENIT and KHATT), two English (IAM and CVL), one Dutch (Fire maker), one Portuguese (BFL), one Chinese (CERUG-CN), one French (LAMIS-MSHD), and one hybrid-language (ICDAR2013).</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Offline writer identification using LCP-IWSL and LBLP descriptors for detecting texture information

  • 期刊论文

  • 发表日期：

    2025年3月20日

查看全文
  • Tayeb Bahram

摘要

Writer identification problem has become prevalent topic in the field of handwriting biometrics. This paper proposes a new approach for writer identification of handwritten documents. The scanned handwritten images are represented by writing descriptors. In order to explain the intra- and inter-writer variability by computing the similarity measurements, two statistical texture descriptors for characterizing the writers’ handwriting styles has been analyzed. Therefore, the joint probability of LCP-IWSL and LBLP on various pixels is calculated. After that, a combination of these descriptors is performed using the sum rule fusion. Identification is carried out using k-Nearest Neighbors and the Chi-Square distance with the simplest kind of cross validation: Holdout. The proposed scheme achieves interesting performances, according to the experimental results on nine well-known handwriting databases, including two Arabic (IFN/ENIT and KHATT), two English (IAM and CVL), one Dutch (Fire maker), one Portuguese (BFL), one Chinese (CERUG-CN), one French (LAMIS-MSHD), and one hybrid-language (ICDAR2013).