IRJET - An Enhanced Approach for Extraction of Text from an Image using Fuzzy Logic

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 03 | Mar 2020 www.irjet.net p-ISSN: 2395-0072
© 2020, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 3107
An Enhanced Approach for Extraction of Text from an Image using
Fuzzy Logic
K. Chandra Sekhar1, MD. Mehar Naaz2, K. Haritha3, N. Roshitha Varma4, T. Sivajee5,
P. Sri Ram Naidu6
1Professor, Dept. of Computer Science and Engineering, Anil Neerukonda Institute of Technology and Sciences,
Andhra Pradesh, India
2,3,4,5,6Student, Dept. of Computer Science and Engineering, Anil Neerukonda Institute of Technology and Sciences,
Andhra Pradesh, India
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract – This paper represents the approach for
extracting the text from an image with the method called
“Fuzzy Logic. The role of text recognition is to recognize the
text from an image and then to extract the text in the editable
format, so that people can use the text fortheirfuturepurpose.
People capture lot of useful information which may have
essential textual data to store and may also edit this data.
Fuzzy logic identifies whether thecharacterorthetextpresent
in an image matches with the trained data or not. The image
captured by the user undergoes pre-processing and then
converted into binary image. The text in an image is localized,
segmentation process takes place to extract each character
from image and then characters are identified which are
matching the trained data using Fuzzy Logic.
Key Words: Text Recognition, Text Extraction, Fuzzy
Logic, Pre-processing,Segmentation,Featureextraction.
1. INTRODUCTION
Textual data present in the images contain useful
information for indexing and automatic annotations.
Extraction of this useful information involves text detection,
localization of text, classification, and then recognition of
text. Fuzzy logic determines the degree of truth values. This
logic helps to identify and match the characters accurately
with trained data.
1.1 Image Processing
Image processing is analysis and manipulation of a digitized
image, so as to enhance its quality with the help of
mathematical operations by using any kind of signal
processing where the input isa picture oranimageoravideo
frame. The output of image processingwillbeeitherapicture
or set of characters or parameters associated with the given
input image. This is a set of computational techniques for
analyzing, enhancing, compressing and reconstructing
image. There are different techniques for processing an
image like optical methods, fuzzy techniques, digital
processing, linear scaling.
Image processing generally involves three steps:
 Importing and Loading the image by using image
acquisition tools.
 Analyzing and manipulating image to extract the
information.
 Output the result. The result might be the image ora
picture altered in some way or it may be a report
based on analysis of the image.
1.2 Text Recognition and Extraction Model
The method of text extraction involves mainly two
processes, namely- Text Detection and Text Recognition.
Text detection is a way of localizing various regions from an
image which contains text in it. This helps in removal ofnon-
text regions which behave as noise while extraction of
desired text. Whereas, Text recognition is a process of
converting pixel-based text i.e. image text to readable code.
This recognizes the text from an image by undergoing
several steps such as, pre-processing, segmentation, feature
extraction, classification and post-processing. Later,thetext
extraction is done by comparing the segmented characters
with the trained data with the help of fuzzy logic.
2. FUZZY LOGIC
Fuzzy logic is a form of logic that is used in some expert
systems and in various applications of Artificial Intelligence.
This was originated in 1965 by “Lofti Zadeh”. Fuzzy logic isa
branch of logic which uses degrees of membership in the
form of sets rather than the strict true/false memberships.
The classical logic depends only on the Boolean values 0or1
and this depends upon a lot of relationships, while there are
various relationships where the position that it can be
considered as partly true or as partly wrong at the same
time. Therefore, Fuzzy logic determines the degree of
affiliation, which is extent of grades between right and
wrong. This can manage the vagueness and ambiguity very
efficiently. It has the power to perform reasonable and
meaningful operations. In the process of extraction of text
this logic plays a vital role by takingthesegmentedcharacter
and comparing this segmented character with the trained

character datasets. After comparing, it finds the accurate
matching value of the character. If character matches
accurately with trained data then the character is extracted
and displayed on the digital screen. If the segmented
character matches with the trained dataset character then
the status becomes ‘1’ and the character is printed,
otherwise status becomes ‘0’ and the processcontinuestill it
finds the matching character.
3. DESIGN AND IMPLEMENTATION
Fig -1: Architecture of the proposed system.
The algorithm involves few steps:
1. Load the original (RGB) image and convert it into
grayscale image.
2. The image undergoes pre-processing to enhance
and remove the noise from an image.
3. The words in the image are classified and localized
and then they are segmented according to the
boundary values.
4. The segmented characters are compared with the
trained dataset with the help of fuzzy logic to find
the highest matching criteria.
5. If the segmented character has highest matching
value with trained data then the character is
displayed or else the same process continues till it
finds the highest matching value i.e. nearer to 1.
6. The characters are displayed in an editable format.
3.1 Pre-processing
The main purpose of pre-processing is to enhance the visual
look of the image. People generally take the picture of the
required document to be extracted as text, but they might
contain some noise. The imagecapturedbytheusermightbe
blurry, noisy, and may be of low resolution. Thus, this
process is used to remove the clamor or noise, stabilize the
intensity of the image and to clear the artifacts. So as to
improve or enhance the quality of the input image given by
the user, few operations are performed in this stage. They
are noise removal, normalization, binarization etc.
3.1.1 Noise Removal
The sources of noise arise in an image due to image
acquisition i.e. digitization and transmission. There are
generally four types of noises, namely: Gaussian noise, Salt
and pepper noise, Speckle noise, Uniform noise. When the
images are sent over different channels, they are prone to
corruption with noise because of the noisy channels. Thus,
filters are required to remove the noise from the images
captured by the user before processing. There are many
kinds of filtering such as linear smoothing filter, median
filter, wiener filter and Fuzzy filter. Thethreeprimaries(R,G
and B) are done separately in the filtering. The three R, G, B
filters are followed by the gain to compensate the reduction
of noise resulting from the filter. Thesefilteredprimariesare
combined to form a filtered colored image. This process is
shown as below:
Fig-2: Filtering
Median filter is utilized for enhancing the image,
Filtered_Image= Median _Filter(Original_Image, Filter_Size)
3.12 Normalization
The range of pixel intensity values are changed in the
process of normalization. In general, normalization means a
mechanism to bring something to normal condition. The
normalized image has mean =0 and variance =1. The range

of values of the image represented between 0 and 255. The
normalization of an image is performed as,
Output_channel = 255 * (Input_channel - min) / (max-min)
If a grayscale image is used then only one channel needs to
be normalized. However, if we are normalizing an RGB (3
channels) we should normalize for each channel using the
same criteria.
3.2 Segmentation
The process of partitioning a digital image into several
multiple segments or image objects is known as
segmentation and sometimes also referred as object
detection. This process simplifies and changes the features
or representation of an image so that analysis becomes easy
and also to make it more meaningful. This segmentation is
done in three ways, they are line segmentation, word
segmentation, character segmentation. Some of the deep
learning architectures are used for segmentation, one of the
architectures is CNN i.e. Convolution Neural Network.
Segmentation of image using CNN involvesfeedingsegments
of an image as input to a convolution neural network, which
indicates the pixels. The entire image in CNN can’t be
processed. It scans the image or picture, looking at a small
“filter” of several pixels each time until it has mapped the
entire image.
Step-by-step process of how this works:
• Take the weight matrix.
• On top of the image put the weight matrix.
• Element-wise multiplication is performed and output is
considered.
• The weight matrix is moved as per the chosen stride.
• Convolve until all the pixels of the input are used.
Each and every segmented character is placed in the form of
matrix. These matrices are compared with the trained
dataset matrix so as to identify the character.
3.3 Feature Extraction and Classification
Features are the unique signatures of an image or an image
defined with unique properties. Feature extraction is
basically related to dimensionality reduction and is used for
extracting characteristics of an input image. This efficiently
describes components of image as a compressed feature
vector. This kind of approach is useful when size of image is
very large. When the pre-processing and segmentation on
the image is done, some feature extraction technique is
applied to the segmented characters to acquire features,
which is then followed by classification and post processing
techniques. Some of the techniques like statistical and
geometrical features are used in feature extraction process.
The process of recognizing and extraction of character is
divided into 2 stages: Feature selection and classification.
The main aim of features selection is to select a subset of
input variables by cutout features with weakly or no
predictive information while maintaining or performing
classification accuracy, whereas Image classification is
acceptance of the input image and the extracting characters
and assigning them to correct class or a category.
4. RESULT
Fig:3(a) Extracted text from input image
Fig:3(b) Extracted text from input image
5. CONCLUSION
Applications need several kinds of images as sources of
information for elucidation and analysis. The characters
which are identified are classified into meaningful word or
sentence. When an image is transformed from one from to
another by digitizing, scanning, processing is done through
tesseract by storing the identified data and this identified
fata is compared with the trained data using the Fuzzy logic.
Therefore, the output image has to undergo a process called
image enhancement, which has a groupofmethodsthatseek
to develop the visual presence of a picture or an image. In
this paper, we have successfully able to extract the text from

the image of any kind of font style and size with the help of
fuzzy logic. The intend of Fuzzy rules aretheattractiveresult
to improve the quality of edges and find the accurate
characters from the image. This paper will act as a good
survey of researchers who have begun work in the field of
fuzzy character recognition.
REFERENCES
[1] L. Neumann and J. Matas. A method for text localization
and recognition in real-world images. In ACCV, pages 770–
783, 2010.
[2] Reza Sarshogh and Keegan Hines,” Computer Vision
Methods for Extracting Text from Images”, Capital One
Tech.
[3] T.Som, Sumit Saha,"Handwritten Character Recognition
Using Fuzzy Membership Function", International
Journal of Emerging Technologies in Sciences and
Engineering, Volume 5, December 2011
[4] L. A. Zadeh. Fuzzy sets, Information Control 8 (1965) 338–
353.
[5] Gur, Eran, and ZeevZelavsky, “Retrieval of Rashi Semi-
Cursive Handwriting via Fuzzy Logic,” IEEE
International Conference on Frontiers in Handwriting
Recognition (ICFHR), 2012
[6] Thomas Natsvhlager, “Optical CharacterRecognition”,A
Tutorial for the Course Computational Intelligence.
[7] Andrei Polzounov,ArtsiomAblavatski , Sergio Escalera,
Shijian Lu, JianfeiCai “Wordfence: Text Detection In
Natural Images With Border Awareness”
[8] D.Trier ,A.K.Jain ,T.Taxt , “Feature ExtractionMethodfor
Character Recognition-A Survey” ,Pattern Recognition

IRJET - An Enhanced Approach for Extraction of Text from an Image using Fuzzy Logic

More Related Content

What's hot (19)

Similar to IRJET - An Enhanced Approach for Extraction of Text from an Image using Fuzzy Logic (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET - An Enhanced Approach for Extraction of Text from an Image using Fuzzy Logic