Text Extraction from Image and its

Language Translation System Using OCR

Project Report Submitted for

Submitted by
Darshan Dharmik

Seat Number: T184011

SCHOOL OF COMPUTER ENGINEERING AND TECHNOLOGY

Pune - 412105, Maharashtra (India)

2018-19
 CONTENTS

Abstract 1

1. Introduction

1.1 Motivation 1

2. .Literature Survey 2

3. System Design

3.1 Hardware Requirements 2

3.2 Software Requirements 2

Implementation Details 3
4.

Duration of Project 4
5.

Cost of the Project 4

7. Conclusion 5

References 5
Abstract

The language barrier among tourists was one of the major difficulties when travelling. The
tourists can rely on mobile phone for travelling purposes. Mobile dictionary is used for word
search for a better travelling guide. Android platform has gained popularity in recent years in
terms of market share and number of available applications. Android makes it easier for
consumers to get and use new content and applications on their Smart phones. This project aims
to develop an Android- based app for text translation application that is able to recognize the text
present on sign boards, images taken with a mobile phone camera, translate the text from
German to English, and display the translation result back onto the screen of the mobile phone.

1. Introduction
1.1.Motivation
People travelling to different places find it difficult to communicate with local people as they do
not know the language. They are unable to interpret the words written on any board or banner. So
there is a need to develop text information extraction systems that can identify and recognize text
that is contained in the images. The project was selected keeping in mind the need to develop an
Android App which will extract the text from images in cases where the user cannot manually
enter the characters. After the method of text extraction, the characters are translated in the user
understandable language. Thus the user will have faster access to the unknown language and will
be able to interpret it.

The motivation of a real time text translation mobile application is to help tourists navigate in a
foreign language environment. The application we developed enables the users to get text
translate as ease as a button click. The camera captures the text and returns the translated result
in real time.

-1-|Page
 2. 1Literature Survey
Ayatullah Faruk Mollah, Nabamita Majumder, Subhadip Basu, and Mita Nasipuri (2011) [1]
presents a complete Optical Character Recognition (OCR) system for camera captured image
textual documents for handheld devices. Firstly, text regions are extracted and skew corrected.
Then, these text regions are binarized and segmented into lines and characters. Characters are
passed into the recognition module. Pranob K Charles, V.Harish, M.Swathi (2012) [2] describes
the techniques for converting textual content from a paper document into machine readable form.
The computer actually recognizes the characters in the document through a revolutionizing
technique called Optical Character Recognition. Chirag Patel, Atul Patel, Dharmendra Patel
(2012) [3] recognize the characters in a given scanned documents and study the changes in the
Models of Artificial Neural Network. It describes the behaviors of different Models of Neural
Network used in Optical Character Recognition. Padwal, KaranS. Punjabi, and Prof. Gandhi S.
Gurjar(2014) [4] presents a simple, efficient and minimum cost approach to construct OCR for
reading any document that has fix font size and style or handwritten style. In this the systems
have the ability to yield excellent results. It is mostly used with existing OCR methods,
especially for English text. . Sravan, Shivanku Mahna, Nirbhay Kashyap (2015) [5] explains that
problems being faced by the developers in using OCR as a technology on a large scale and give
the solution to that problem. This system provides many features that require no typing, editing
raw data, quick translation, and memory utilization. Surabhi Dusane, Monica Ahuja, Rucha
Ghodke & Prathamesh Kothawade (2016)

2.2Market Survey

Our project will mainly be helpful for the travelers, who travels to different countries, but don’t
know the local language of the particular country. Let’s say you are heading to China, Korea or
Japan and want an app to read the signboards and menus for you? Our system will fill your need
by giving you the English translation for street signs, menus or anything your smartphone
camera can point at. Moreover, even the people who wants to learn new language, or want to
know meaning of any foreign word, can take help of our project.

-2-|Page
3. System Design

3.1.Hardware Requirement:
 Ubuntu 16.4
 Ram- 4Gb
 Hard disk- 500GB

3.2. Software Requirement:

 Python 2.7

 Python-Libraries, Open-cv , Lang detect, pytesseract

 Language translation – translate Py library

3.3 Design Considerations:

 Camera: The user will click the picture of the text that needs to be understood.
 Gallery: If the image was taken previously, the user can select the image from gallery of the
mobile.
 Crop function: After image is given as input, the user is given an option to crop only the
required word from the whole text. Only that word would be extracted
 Translate: After getting the extracted text, the user can press the translate button to get the
meaning of extracted text.

-3-|Page
4. Implementation Details

Optical Character Recognition (OCR), is a technology that enables you to convert different types of
documents, such as scanned paper documents, PDF files or images captured by a digital camera
into editable and searchable data. It is the mechanical or electronic conversion of images of
typewritten or printed text into machine encoded text. Images captured by a digital camera differ
from scanned documents or image-only PDFs. They often have defects such as distortion at the
edges and dimmed light, making it difficult for most OCR applications, to correctly recognize the
text. It offers a range of features to improve the quality of such images, providing you with the
ability to fully use the capabilities of your digital devices.

After extracting the words from image by using the Optical Character Recognition (OCR) Engine,
those words are translated into known language, to do this the Language Translator service is used.
This is a free service. It provides many libraries for translation. The first thing to remember is that
translation is the transfer of meaning from one language to another.

Input:

 The captured image.

Assumptions:
 The text is in normal font i.e. times new roman or Calibri.

 There are no much noise in the image

Output:

 Translated text.

-4-|Page
 5. Duration of the project

The duration of the project was 12 days i.e. 38 hours.

Activity Duration(days) Duration(hours)

Decide the project 2 8

Literature Review 2 8

Coding 6 18

Testing 2 4

Total 12 38

6. Cost of the project

Cost
Components (INR)

Laptop 30,000

Internet 2500

HR Cost 20,000

Total 52,500

-5-|Page
7. Conclusion
We have achieved an Android based application for real-time text extraction, recognition and
translation. The average correct character-recognition rate is above85%.
Following work needs to be done in order to drive our application into a commercial product:

•Further optimize text extraction algorithm to in-crease the processing speed to real-time (multi-
thread processing structure in Android Java code);
•Adapt text extraction algorithm to smaller and denser text;
•Local Otsu’s method to adapt non-uniform back-ground of the scene;
•Localize the OCR, text correction and text translation algorithms onto the mobile device;
•More language translation selections for the user;

8. References

[1] Yassin M.Y.Hasan and Lina J.Karam,Morphological Text Extrac-tion from Images.IEEE
Transaction on Image Processing Vol.9No.11, Nov 2000
[2] Dimitrova, L. Agnihotri [1999], “Color super-histograms for video representaion. In: Proc.
Of the International Conference on Image Processing.
[3] http://code.google.com/p/tesseract-ocr/
[4] http://norvig.com/spell-correct.html
[5] http://austingulati.com/2009/07/google-translate-php-api/
[6] Farshad Ghazizadeh,Optical Character Recognition. US Patent:5,007,809.
[7] Huiping Li, David Doermann and Omid Kia,Automatic TextDetection and Tracking in
Digital Video.IEEE Transaction onImage Processing Vol. 9 No. 1, Jan 2000
[8] C.P. Sumathi, T. Santhanam, G.Gayathri Devi, “A Survey On Various Approaches Of text
Extraction InImages”, International Journal of Computer Science &Engineering Survey
(IJCSES). Vol.3, August 2012, Page no. 27-42.

Page 6 of 9