INTERNATIONAL JOURNAL OF PROGRESSIVE RESEARCH IN SCIENCE AND ENGINEERING, VOL.3, NO.05, MAY 2022.

Text Recognition Images using OCR

Karthikeyan G 1, Bharanidharan G1, Jeevanandham D1, Balaji B G1
1Student, Department of Computer Engineering, Sri Eshwar college of Engineering, Coimbatore, India.
Corresponding Author: [email protected]

Abstract: - In the current environment, visualizing images from photos could be a critical procedure. The storing of data found in
paper documents in digital format has exploded in recent years. This aids within the storage of information, is straightforward to
store, and enables the retrieval of information when needed. Pre-processing, classification, extracting features, classification, and
post-processing are the stages of text recognition. The front processing stage entails a spread of operations that are required to
convert a color image into a dynamic image with text separated from the background. The separation stage aids in character
separation. To help in word recognition, a feature called "Extract" allows you to extract the foremost relevant information from a
picture. The classification process aids within the identification of text by using well-defined principles. The information is then
subjected to post-processing so as to cut back mistakes. During a few operating systems, text recognition is critical. The text
recognition module is discussed during this work, as are several applications of text visualization in photographs. This paper
examines related activities.

Key Words: — Text recognition, preprocessing, classification, post processing.

I. INTRODUCTION Optical Character Recognition (OCR) is one among the

foremost widely used messaging technologies. It scans photos
In recent years, text recognition has gained lots of traction and converts them to text. Text recognition could be a time-
because it has expanded into new areas like automatic license consuming process that needs recognizing text in an
plate reading and sign boards. during this day and age, exceedingly type of fonts, styles, and backgrounds. thanks to
everything and everybody has gone digital. Nowadays, the changes in character size, form, and spacing between
photographs or scanned words are wont to transmit the bulk characters that change from person to person, recognizing
of the data. At any given time, there's an enormous volume of handwriting is considerably harder. As a result, an automatic
information being saved and retrieved. the requirement to text recognition system must be developed that may detect an
keep up and access information comprising texts in an element of existing text in a picture or scene and convert it to
exceedingly simple and quick manner drives the sector of text a machine-readable format. Taking an image of the relevant
recognition. Scanning turns information into a picture and document, processing it to find the part you would like, then
prevents scanned information from being reused within the sorting it to extract the prevailing text content are all parts of
sort of text, making it one in all the best ways to transfer the text recognition process. the various phases of text
information from paper or books. As a result, upgrading the recognition from photographs are discussed during this work.
tools is required so as to rework them into an ordered format.
The goal of this work is to appear at the various processes II. LITERATURE
within the text recognition process so as to convert text images
into structured text. This section gives a fast review of this state of text
recognition research. Text recognition has been around for a
protected time.
Manuscript revised May 12, 2022; accepted May 13, 2022. Colorful backdrop images are considered and
Date of publication May 15, 2022.
defined earlier for the Tesseract Optical Character
This paper available online at www.ijprse.com Recognition (OCR) engine's advanced processing in [1. By
ISSN (Online): 2582-7898; SJIF: 5.59 separating truth image into k images, the initial text is
separated from the colorful background. Then sort
by watching the image with the text on that. compared to the

KARTHIKEYAN G., et.al: TEXT RECOGNITION IMAGES USING OCR 57

 INTERNATIONAL JOURNAL OF PROGRESSIVE RESEARCH IN SCIENCE AND ENGINEERING, VOL.3, NO.05, MAY 2022.

Tesseract OCR's performance with hiring pre-processing, III. TEXT RECOGNITION MODULE
there was a 20% improvement.
A Number of jobs should be done by the text recognition
Akopyan, O.V. Belyaeva, and T.P. Plechov's work are
module. The module accepts a text as input. The output
featured overagain.
module generates textual data in a machine-readable format.
D.Y. Turdakov [2] is made on a duplication pipeline that Preprocessing, sorting, extracting features and separating
extracts text from a spread of social media photos. Their should all be done by the text recognition module. Figure 1
major task is to divide the included photos into different depicts the many functions involved in text recognition.
classes, with the primary processing allotted per the A scanned document is often in the form of an image. The first
classifications. After that, the OCR engine is employed to step is to process the image so that it can be converted into a
recognize text. This work makes use of a database generated format that can be used for further processing. Audio or
from social media. OCR may be a technique for identifying warped text images are possible. The image is enhanced in this
the text portion of a picture. The authors of [3] presented a stage by eliminating the sound and converting it to binary. The
way for extracting text from a scanned document. The sound in the image plays an important part in correctly
partition Otsus technique and also the skew detection Hough understanding the text. The elimination of audio improves the
transform were employed during this study. The characters odds of proper text visualisation and results in more precise
are identified using the OCR method. They test and validate output. To remove audio, many filters, such as the Gaussian
the suggested technique on a range of photos from various filter and the definition filter, can be employed. Then, to
sources. The accuracy rate was 93 percent. validate the similarity, a practise is made, followed by a binary
Karthick, K.B. Ravindrakumar, R. Francis, and S. Ilankannan to convert a grey image to a binary image.
[4] went over the various processes of text acquisition in great
depth, highlighting the assorted strategies that were used.
They also place a premium on handwriting recognition, which
is one in every of the foremost difficult skills to master. in line
with their research, the most effective results could also
be obtained by reducing calculation time and improving
character identification by distinguishing between characters
from multiple languages.
Anupriya Shrivastava, Amudha J., Deepa Gupta, and Kshitij
Sharma [5] developed a Convolutional Neural Network and
a remembering Programme. the improved model recognises
text in images that are horizontal, curved, or vertical. The
model is formed of four sections. The element is removed at a
lower level within the first portion. The distributed
convolution approach is employed within the half to extract
high-quality features. The third party is unconcerned with the
minor details. The fourth section anticipates the letter
sequence.
In their work, Pratik Madhukar Manwatkar and Dr. Kavita R. Fig.1. depicts a variety of text recognition tasks.
Singh [6] discussed numerous methods for extracting
Individual characters are separated using a separation
characters from photographs. In their paper, they also
technique after the processing is completed. The feature
outline the essential structure of a text recognition
removal phase is then used to restore critical data to the raw
system supported photographs. They also talked about a way
data.
to extract text from a scanned image employing a series of
image processing techniques. In their article, they also think Segregation can be accomplished using a variety of
again the various application fields. methodologies, including Principle Component Analysis
(PCA), Linear Discriminate Analysis (LDA), Independent

KARTHIKEYAN G., et.al: TEXT RECOGNITION IMAGES USING OCR 58

 INTERNATIONAL JOURNAL OF PROGRESSIVE RESEARCH IN SCIENCE AND ENGINEERING, VOL.3, NO.05, MAY 2022.

Component Analysis (ICA), Chain Code (CC), Histogram, Text recovery also makes a large number of books available
and so on [6]. for storing and sharing on the internet information.
The next stage is to divide, which entails recognising each Text recognition also aids industry automation by allowing
letter and allocating it to the relevant letter group, resulting in for automatic label and number reading. Visual recognition
the text being converted to a machine-readable format. For has grown in popularity as technology has advanced, and it
this objective, several separators based on artificial neural now has a wide range of applications in practically every
networks (ANN) and vector support machines (SVM) may be sector.
used.
The term "post processing" refers to the act of saving known
text in a format that can be used for subsequent processing.
3.1 System Specification:
Software Used:
 Operating System: Windows 7 / 8/ 10
 Language : Python
 IDE : Anaconda,
Notebook
Hardware Used:
 Processor : Intel core i3
Fig.2. Results
 Ram : 8 GB
4.1 Summary
 Hard Disk : 120 GB
This paper presents a brief summary of the various steps used
IV. APPLICATION to identify text from images. The work done in this field is
also briefly discussed. An update to the basic model of the text
A number of text-capture applications have gained popularity recognition system is also provided describing the flow of text
in recent years. In a scanned document, the automatic text recognition from images. Finally, a discussion of the various
detection is excellent. Recent technological advancements areas in which text recognition can be used.
have made text-recognition algorithms possible. Text
recognition allows for automation in a wide range of V. CONCLUSION
industries. Automatic licence plate reading in paid areas;
automatic reading of leaf signature checks; image tagging Image processing is used to extract important information
and group data analysis are all possible applications. from an image. The image processing algorithm is first used
and those images are fed in different ways to get accurate
In the healthcare industry and offices, document recognition
results. One of the methods used is CNN; the image process
also provides automatic storage and access to large records. It
is to feed the convolution layers to get the output prepared and
allows you to create a site where text can be readily searched
the other way around is tessaract Tessaract is a good trading
and indexed. The information that has been created may be
engine about improving accuracy. Character is seen
retrieved and modified with little effort, which saves a lot of
effectively and reliably using LSTM algorithms. Comparative
paper work. Visually challenged people can benefit from
image analysis in different entropy showed that one can
automatic text recovery with voice aids.
achieve output near perfection and output will vary from one
It can also be used to show a portion of a video's text. period to another. The result of each stage is therefore
Intelligent transportation systems can benefit from automatic observed and is subject to OCR and the best output is
text recognition. It can also be used to verify passports and considered a result.
other information at airports. Automatic data entry for
business papers is also possible with text acquisition.

KARTHIKEYAN G., et.al: TEXT RECOGNITION IMAGES USING OCR 59

 INTERNATIONAL JOURNAL OF PROGRESSIVE RESEARCH IN SCIENCE AND ENGINEERING, VOL.3, NO.05, MAY 2022.

REFERENCES
[1]. Matteo Brisinello, Ratko Grbi, Dejan Stefanovi and Robert
Pekai- Kova, Optical Status Recognition in Colourful Photo
Gallery, 2018, IEEE 8th International Conference on
Consumer Electronics - Berlin (ICCE-Berlin).
[2]. M.S. Akopyan, O.V. Belyaeva, T.P. Plechov and D.Y.
Turdakov, Text recognition in photos from social media,
2019, Ivannikov Memorial Workshop (IVMEM).
[3]. Neha Agrawal, Arashdeep Kaur, Algorithmic Method of
Text Visualization from Printed / Typed Text Images, 2018,
8th International Conference on Cloud Computing, Data
Science and Engineering.
[4]. K. Karthick, K.B. Ravindrakumar, R. Francis, S.
Ilankannan, Steps Involved in Text Recognition and Recent
Research in the OCR; Research, International Journal of
Modern Technology and Engineering (IJRTE) ISSN: 2277-
3878, Volume-8, Issue-1, May 2019.
[5]. Anupriya Shrivastava, Amudha J., Deepa Gupta, Kshitij
Sharma, In-Depth Learning Model, 10th ICCCNT 2019 July
6-8, 2019, IIT - Kanpur, Kanpur, India.
[6]. Pratik Madhukar Manwatkar, Dr. Kavita R. Singh,
Technical Review of Visualization of Photography, IEEE
Sponsored by 9th International Conference on Intelligent
Systems and Control (ISCO), 2015.

KARTHIKEYAN G., et.al: TEXT RECOGNITION IMAGES USING OCR 60