Please use this identifier to cite or link to this item:
http://hdl.handle.net/10603/226724
Title: | A Novel Approach for Text Recognition in Devanagari Script |
Researcher: | Vaidya Madhav V. |
Guide(s): | Joshi Yashwant V. |
Keywords: | Engineering and Technology,Engineering,Engineering Electrical and Electronic |
University: | Swami Ramanand Teerth Marathwada University |
Completed Date: | 26/03/2018 |
Abstract: | Off-line handwritten Devanagari character recognition (DCR) is a very hard pattern newlinerecognition problem and has of practical importance for digital India concept. Two popular newlineapproaches are to extract features holistically from the character image or to decompose newlinecharacters structurally into component parts most usually strokes. Different newlinetechniques are explored in the field of Devanagari Text Recognition in the context of classifying newlinenumerals and characters from handwritten Marathi documents. There is scope newlinefor improvement in every stage of DCR system like preprocessing, segmentation, feature newlineextraction and classification. newlineFeature extraction can be done holistically on the character image or decomposition newlineof character image can be done to divide the image structurally into different components newlinelike strokes, zones etc. First a complete OCR system for handwritten Devanagari document newlinehas been studied and segmentation techniques have been investigated. A novel newlineapproach is proposed for calculating summation vectors as a feature vector for character newlineimage based on horizontal, vertical and diagonal parts illustrating statistical distribution. newlineDiscrete cosine transform is used for dimensionality reduction. Likewise separation newlineof machine-printed and handwritten documents is performed. newlineThe isolated characters approaches for segmenting individual characters from whole newlinedocuments has been investigated, considering the Shirorekha with words. Shirorekha newlineis used to determine the upper strip, middle strip, and lower strip but we are not eliminating newlinethe Shirorekha because the database created also have Shirorekha in each vowels newlineand consonants. The proposed method is 100% efficient for the segmentation of line and newlineword. The segmentation of characters from word is critical, as single word may contain newlinecomposite characters i.e. combination of vowels and consonants. The histogram projection newlinemethod has been used for segmentation which gives 99% accuracy in result of newlinesegmentation of line and segmentation of word. newlineIn one of the proposed r |
Pagination: | 122p |
URI: | http://hdl.handle.net/10603/226724 |
Appears in Departments: | Faculty of Engineering |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
01_title.pdf | Attached File | 114.3 kB | Adobe PDF | View/Open |
02_certificate.pdf | 32.31 kB | Adobe PDF | View/Open | |
03_abstract.pdf | 42.17 kB | Adobe PDF | View/Open | |
04_declaration.pdf | 25.97 kB | Adobe PDF | View/Open | |
05_acknowledgement.pdf | 19.21 kB | Adobe PDF | View/Open | |
06_content.pdf | 57.03 kB | Adobe PDF | View/Open | |
07_list_of_tables.pdf | 39.81 kB | Adobe PDF | View/Open | |
08_list_of_figures.pdf | 56.67 kB | Adobe PDF | View/Open | |
09_list_of_algorithms.pdf | 39.72 kB | Adobe PDF | View/Open | |
10_chapter 1.pdf | 538.88 kB | Adobe PDF | View/Open | |
11_chapter 2.pdf | 161.01 kB | Adobe PDF | View/Open | |
12_chapter 3.pdf | 1.54 MB | Adobe PDF | View/Open | |
13_chapter 4.pdf | 526.87 kB | Adobe PDF | View/Open | |
14_chapter 5.pdf | 217.9 kB | Adobe PDF | View/Open | |
15_chapter 6.pdf | 255.51 kB | Adobe PDF | View/Open | |
16_conclusions.pdf | 60.71 kB | Adobe PDF | View/Open | |
17_bibliography.pdf | 115.31 kB | Adobe PDF | View/Open |
Items in Shodhganga are licensed under Creative Commons Licence Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0).
Altmetric Badge: