Article Preview
TopIntroduction
Optical recognition system (OCR) is an important technique for processing images to turn typed or manuscript records into editable text formats. The accuracy of the OCR depends upon the scanned quality of the image. The OCR for manuscript records is generally more complex than printed records because in printed records, words or characters follow some strict rules, and they will be in standard formats. The hand written records mostly (of different users or a particular user) does not have a unique format and style. The manuscript records of a particular user also vary based on environment, the type pen, strokes, mindset, mood etc. Further there will be large variations of manuscript records between individuals. That’s why Slope and slant estimation and correction are the crucial preprocessing steps of manuscript records. Typically, the methods of slope estimation measure the angle between the horizon line and the line a character or term is aligned along. For slope correction, an image of the de-slope is produced. The method of calculating slant angles calculates the angle between the vertical axis and the word's most dominant vertical stroke. The shear transformation is applied to de-slant the manuscript word. In the literature [1 – 23], many slant detection methods were presented on many European languages. However, only a few works are reported on Indian languages and especially on Dravidian-based languages; one or two works were reported. In(Vinciarelli & Luettin, 2001), in cursive measured dimensions, Alessandro Vinciarelli and Juergen Luettin proposed new techniques of slant and gradient reduction. Three model for predicting local slants for manuscript words in English were suggested by Yimei Ding et al.(Ding et al., 2004). In (Papandreou & Gatos, 2012) A. Papandreou and B. Gatos proposed a unique paradigm for the estimation of word slant by non-horizontal parts analysis of the characters and core region information. By combining the initial and its inversion, Farzad Nadi et al.(Nadi et al., 2013) suggested a slant correction system for Persian manuscript digits and words. A. M. Hafiz and G. M. Bhat proposed nonuniform slat estimation method for slant estimation in (Hafiz & Bhat, 2016). Proposed slope and slant correction of offline manuscript text words in(Das Gupta & Chanda, 2014) Jija Das Gupta et al.(Dingt, 2000) Yimei Ding et al.(Dingt, 2000) proposed an iterative block based approach for slant approximation and correction. Karim Faez et al. Majid Ziaratban •, (Ziaratban & Faez, 2009) proposed nonuniform slant estimation and correction for Farsi/Arabic manuscript words using near-vertical strokes. In (Taira et al., 2004) Eiji Taira et al., proposed a nonuniform slant correction technique by performing a dynamic program based algorithm from the sequence of local slant angles. Ishaan Agrawal et al., (Agrawal et al., 2014) presented slant estimation using Radon transform and Hough transform. In the literature, Hough transforms (Agrawal et al., 2014) Correct the angles of a manuscript word image's slope and slant. The major disadvantage is the computational complexity and space complexity of this method (Agrawal et al., 2014), and it's directly proportional to the size of the word image's data pixels. Numerous methods have been proposed in the literature to address the above issues in slant correction. Javad Sadri et al. proposed slant estimation using geometric features of connected components (Sadri et al., 2010). In (Manisha et al., 2016) Ch.N.Manisha et al. proposed slant correction of manuscript Telugu isolated characters and words using vertical starting and end coordinates. In (Ali, 2012), Mohamad introduced the de-slanting method for Arabic scripts using projection profile and Winger-Ville distribution. In (Virajitha et al., 2012) authors proposed method for slant correction of cursive manuscript words. Using affine transformation authors in (Fitrianingsih et al., 2016) proposed for cursive manuscript records. In (Mahanta & Deka, 2013) L.B.Mahanta proposed slant detection and correction for a manuscript signature verification system. In (Naik & Patel, 2014), the authors suggested a de-slant method for offline English scripts. In (Kumar & Bharathi, 2012), the authors proposed slant correction for Ancient Tamil manuscript records using contour points accumulation. In (Jagadeesh Kannan & Prabhakar, 2008), the authors proposed slant correction for Tamil records. In (Gupta & Chanda, 2012), the authors proposed two novel methods for slant correction of manuscript records.