Differentiation of the Script Using Adjacent Local Binary Patterns Darko Brodi´ c 1 , ˇ Cedomir A. Maluckov 1 , Zoran N. Milivojevi´ c 2 , and Ivo R. Draganov 3 1 University of Belgrade, Technical Faculty in Bor, V.J. 12, 19210 Bor, Serbia 2 College of Applied Technical Sciences, Aleksandra Medvedeva 20, 18000 Niˇ s, Serbia 3 Technical University Sofia, Department of Radio Communications and Video Technologies, Boulevard Kliment Ohridsky 8, 1000 Sofia, Bulgaria dbrodic@tf.bor.ac.rs Abstract. The paper proposed an algorithm for script discrimination using adjacent local binary patterns (ALBP). In the first stage, each letter is modeled according to its height. The real data are extracted from the probability distribution of the letter heights. Then, the gray scale co-occurrence matrix is computed. It is used as a starting point for the feature extraction. The extracted features are classified according to ALBP. Because of the variety in script characteristics, the statistical analysis shows the differences between scripts. Accordingly, the linear discrimination function is proposed to distinct the scripts. The proposed method is tested on the samples of the printed documents, which include Cyrillic and Glagolitic script. The results of experiments are encouraging. Keywords: Cryptography, Optical character recognition, Script recog- nition, Statistical analysis, Typographical features. 1 Introduction Script identification represents a key step in multi script and multilingual doc- ument image analysis [1]. The existing approaches can be classified in two cate- gories: global and local [2]. Global approaches utilize the statistical feature and frequency-domain analysis of wider regions, i.e. text blocks in document images [3]. Their correctness is highly sensitive to the quality of the document images and to the level of the noise included in these images. Local approaches segment the characters, words or lines as connected components and analyze the feature that lists the black pixel runs [4]. Hence, this type of the analysis is convenient for low quality document images. It is insensitive to noise, too. However, it is more computer time intensive than global approaches. This paper proposes the combination of the local and global approach. It is established by extracting all characters in text, which are classified and subjected to the statistical analysis. In this way, the modeling of document is performed in the manner like the cryptography. Each letter from text document is superseded with the cipher [5]. The cipher is extracted from the probability distribution of G. Agre et al. (Eds.): AIMSA 2014, LNAI 8722, pp. 162–169, 2014. c  Springer International Publishing Switzerland 2014