Segmentation of handwritten words using structured support vector machine
- PDF / 3,639,912 Bytes
- 13 Pages / 595.276 x 790.866 pts Page_size
- 15 Downloads / 249 Views
SHORT PAPER
Segmentation of handwritten words using structured support vector machine Manoj Kumar Sharma1 · Vijaypal Singh Dhaka1 Received: 25 June 2018 / Accepted: 22 August 2019 © Springer-Verlag London Ltd., part of Springer Nature 2019
Abstract Words and characters segmentation is a most indispensable and fundamental task for the handwritten script recognition. However, the complex language structures, deviation in pen breadth and slant in inscription make the feature extraction process very challenging. In this research, a binary quadratic process has been formulated for the word segmentation. It deliberates a co-relationship between the inter-word gap and intra-word gap. The structured support vector machine is used for the experiment. Experimental results of public datasets (i.e., ICDAR2009 and ICDAR2013) show state-of-the-art performance of the designed algorithm. Keywords S-SVM · Inter-word · Intra-word · SURF descriptors, SO
1 Introduction Segmentation of line, word and character of the offline handwritten text is very rigorous in the field of offline handwritten text recognition [1–6]. The uneven writing styles, irregular sizes, slants, inter-gaps and intra-gaps between words and characters have made it very challenging task. In the last decade, some acknowledgeable work has been accomplished in the field of line segmentation [7–9]. However, the progress in the word and character segmentation algorithms is not acknowledgeable. Jindal et al. [10] have proposed a segmentation method for handwritten lines and words, which is based on midpoint detection technique, and accuracy of the results is 95%. Mehdi et al. [11] have improved the accuracy of word segmentation of cursive handwritten texts and also done the comparative analysis of the algorithms works with both bitmap and bitmap data. Jain et al. [12] have proposed a word segmentation technique for the OCR systems. The technique estimated the textual area as a large window. Then after, large * Vijaypal Singh Dhaka [email protected] Manoj Kumar Sharma [email protected]; [email protected] 1
Department of Computer and Communication Engineering, Manipal University Jaipur, Jaipur, India
window is alienated into small subwindows of distinct lines and these lines further segmented into the words as small sub–subwindows. Karmakar et al. [13] have proposed the line and word segmentation of a script using spaces between lines and words. Jindal and Lehal [14] have proposed a line segmentation method for the historical documents which are affected by repeated use and aging problem describing the historical documents which are affected by problems of aging and repeated use by imposing text blocks over the text lines and words. Dahake et al. [15] have proposed a word segmentation method for online handwritten Gurmukhi sentences by considering strokes in which words white space is implicitly known and applied the threshold to segment the words from the text lines by locating the vertical gaps, and then extracting the
Data Loading...
