Proceedings of the Fourth International Conference on Computing Methodologies and Communication (ICCMC 2020)

IEEE Xplore Part Number:CFP20K25-ART; ISBN:978-1-7281-4889-2

Open CV based Information Extraction from

2020 Fourth International Conference on Computing Methodologies and Communication (ICCMC) 978-1-7281-4889-2/20/$31.00 ©2020 IEEE 10.1109/ICCMC48092.2020.ICCMC-00018

Cheques
Aniket Dhanawade, Abhishek Drode, Gifty Johnson, Aadesh Rao and Dr. Savitha Upadhya
Department of Electronics and Telecommunication Engineering, Fr. C.R.I.T.

Vashi, Navi Mumbai, India

dhanawadeaniket72@gmail.com, drodeabhishek1997@gmail.com, gifty.johnson02@gmail.com,
aadeshrao5@gmail.com, savitha.upadhya@fcrit.ac.in
Abstract – In general, bank cheques are used extensively
for financial transactions in various organizations.
Cheques are always verified manually. The traditional
verification process will always include date, signature,
legal information, and payment written on the cheques.
In this paper, extracting the legal information from
captured cheque image is obtained by preprocessing the
image, extracting required information and then
recognizing and verifying the handwritten fields. Image
processing techniques like thinning, median filtering,
dilation, and verification techniques are also employed
in this approach.
Keywords – OCR, TESSERACT, OPENCV
I. INTRODUCTION
Money transfer is a crucial part of all institutions.
There are numerous ways to transfer money like
direct cash, online transactions, demand drafts,
cheques etc. Out of these, money transfer using
cheques is the most widely used method. Every
institution or organization perform confidential
money transfer using cheques or demand draft. The
reason is that use of cheques is more secure as
compared to other methods. But even today, details of
these cheques are manually entered. For examp le, in
schools, when students deposit their cheques, the
officials manually enter their data in an excel sheet.
This task tends to be laborious. The same procedure is
being followed in banks as well, where thousands of
cheques are dealt with each day. The signatures are
verified manually, i.e. the employee visually verifies
if the signature is genuine. The errors occurring in this
type of manual checking is large and time consuming.
So, there is a need for automating the above steps.
Here a system which can scan a cheque and extract all
the necessary informat ion and make a list of all the
transactions happening is proposed along with a
software that verifies signature using an availab le
database is proposed to be imp lemented. II.
LITERATURE REVIEW
Handwritten documents are one of the most reliable
sources of information. One of them is cheque, it is a
written order directing a bank to pay a part icular
amount of money or to withdraw money. This
med iu m of service can be easily cheated, when the
modes of verification are insecure. Therefore, it is
necessary to verify the signature.
There are two types of signature verificat ion methods.
1. Offline Methods
These use static in formation for verificat ion purpose.
Offline signature schemes use the signature as the
input image and usually have noise present in them.
So to eradicate the noise from the signature after
processing, the input image filters such as median
filter are applied. 2. Online Methods
These make use of dynamic informat ion for
verification purpose. Online signature verification is
carried using pressure-sensitive tablets and webcam
that extracts features fro m a signature. [1]
Handwriting Recognition is one of the most active
areas of science where deep neural networks are used.
To train a model of a neural network it takes a large
dataset. The method of identifying handwriting
characters is divided into two systems. [2]
a) Android application: The android program
allo ws the user, by using their mobile device, to select
a text p icture to be recognized. To ext ract the image
informat ion, this image is processed via a python
script running on a server.
b) A server: This is their machine backend.
This server is a machine that can run a python script.
It's needed because an Android smartphone doesn't
have the computational power required to run neural
networks and do image processing operations.
A. OCR (Optical Character Recognition) [3]
OCR is a technology that identifies text within an
image and is used to detect text fro m scanned
documents. OCR processes an image by discovering
and identifying characters. It will expo rt the text o r
convert the characters to editable text d irectly in the

978-1-7281-4889-2/20/$31.00 ©2020 IEEE 93

Authorized licensed use limited to: University of Exeter. Downloaded on June 19,2020 at 13:43:17 UTC from IEEE Xplore. Restrictions apply.
Proceedings of the Fourth International Conference on Computing Methodologies and Communication (ICCMC 2020)
IEEE Xplore Part Number:CFP20K25-ART; ISBN:978-1-7281-4889-2

image. Advanced OCR can export the size, fo rmat

and layout of the text on a page. [3]

B. TESSERACT [4]

Python-tesseract is a Python method for Optical

Character Recognition (OCR). This is used to
acknowledge and interpret the text in the images. It
can read several of the types of images that Pillow,
Leptonica's imag ing libraries support, including jpeg,
png, gif, tiff. If script is used, the detected text will be
printed by Python-tesseract rather than written into a
register [4].
C. OPENCV [5]
OpenCV stands for Open source computer vision.
It is a lib rary of programming functions that aims at
real-t ime computer v ision. It consists of more than
2500 algorith ms wh ich supports computer vision
applications. It has C++, Python, MATLAB and
JAVA interfaces and supports Windows, Linu x,
macOS and it also runs on various mobile
operating systems such as Android, iOS,
BlackBerry.[5]
Fig. 2. Scanned cheque image
B. Cropping of input image
The acquired image is then cropped using the pillow
lib rary. Pillow is a library availab le fo r Python
programming. All cheques have a standard format.
The date, payee, signature, amount is all in the same
place. So manual cropping is perfo rmed i.e. the
coordinates of the image to be cropped is given
manually. This gives us the cropped image of all the
essential data. Fig.3 shows the name of the bank.
Fig.4 shows the amount. Fig.5 shows the account
number. Fig.6 shows the bank address. Fig.7 shows
the date that has been extracted.

III. IMPLEMENTATION

Fig.1 shows the block diagram o f processing of Fig. 3. Bank name

cheques using image processing.

Fig. 4. Amount on cheque

Fig. 5. Account number

Fig. 1. Block diagram Image processing

A. Image acquisition
Fig. 6. Address
In Image acquisition the recognition system takes a
scanned image as an input image. The image should
have a specific format as JPEG. Th is image is taken Fig. 7. Date
through a scanner or a camera. Fig.2 is a scanned
image of a cheque.

978-1-7281-4889-2/20/$31.00 ©2020 IEEE 94

Authorized licensed use limited to: University of Exeter. Downloaded on June 19,2020 at 13:43:17 UTC from IEEE Xplore. Restrictions apply.
Proceedings of the Fourth International Conference on Computing Methodologies and Communication (ICCMC 2020)
IEEE Xplore Part Number:CFP20K25-ART; ISBN:978-1-7281-4889-2

Fig. 8. Detected text from image

C. Pre-processing Fig. 10. Normalized image

Pre-processing is used to generate data that is
suitable for the OCR system to operate precisely.
The OCR system needs clear image with no noise
in them. W ithout pre-processing, the system gives
false output. For e.g. the lines in between the dates
has to be removed to extract the proper date. This
is achieved by performing binary thresholding. A
reference threshold is set and the pixel values
below that are made b lack and the once above the
threshold are made white. This helps in removing
the bars which are at a higher intensity as
compared to the written date.
Pre-processing involves: [6]
a. Noise removing or Normalization
The linear normalizat ion is performed using the given
formula:
In= (I-Min)*(newMax-newMin)/Max-Min+newMin,
Where ‘I’= grayscale image with intensity values
ranging from (Min, Max) and ‘In’= New image
with with intensity values ranging from (newMin,
newMax). [8] b. Binarization [9]
Image binarization is used to reduce the amount of
image information such as color and background of
the image, so that output image is black and white.
This is done by setting a threshold value and
classifying all the pixels above threshold as white and
below threshold as black. Binarized image is shown in
Fig. 11.

In normalization a noise reduction filter such as

med ian filter is enfo rced on the binary scanned
image to remove single white pixels on black
background and vice-versa. Fig. 9 is the original
image and Fig. 10 is the normalized image.
Fig. 11. Binarized image
c. Morphological Operations [10]

The signature area is separated from the background

through segmentation process of vertical and
horizontal pro jection. Morphological operations like
Erosion and Dilat ion are applied to perform th is
operation. Fig. 12 shows cropped image after
thinning. It is obtained by subtracting the hit-and-miss
Fig. 9 Original image [7]
transform of the image and the structured element
from the image.

Fig. 12. Thinned image [10]

978-1-7281-4889-2/20/$31.00 ©2020 IEEE 95

Authorized licensed use limited to: University of Exeter. Downloaded on June 19,2020 at 13:43:17 UTC from IEEE Xplore. Restrictions apply.
Proceedings of the Fourth International Conference on Computing Methodologies and Communication (ICCMC 2020)
IEEE Xplore Part Number:CFP20K25-ART; ISBN:978-1-7281-4889-2

Fig. 13. Csv File

analysis. It helps manage large data systematically. It
creates a .csv file. csv stands for Co mma Separated
Variables. Many columns can be defined and the data
can be segregated. This type of presentation is clear
and precise and is easy for the user to understand. The
obtained csv file is shown in Fig. 13.

After extract ing the cheque features such as the bank

Fig. 14.
Hough Transform
Hough transform of this image is taken. Hough
transform is used to detect forgeries. Hough transform
peaks are used for comparison. Fig. 14 shows the
Hough transform of the input image. [11][12]
Fig. 15 shows the pre-processed date performed using
above steps.
Fig. 15. Pre-processed date
name, cheque amount, account number, bank address
and the date, the next step is to authenticate the
signature which can be done using Neural Network o r
Machine Learning algorithms.
VI. CONCLUSION
A. Conclusion
The objective of the above wo rk presented in this
paper is to make an auto mated system which will
extract information fro m cheques, validate the
informat ion and keep record o f all the info rmation
present that can be used by financial institutes to
reduce human efforts in verification of signature
which is done manually as well as it will keep a
record of all the transactions.

VII. RESULT In this paper bank name, address, account number,

date, amount was extracted fro m the cheque using
A. Image text conversion different libraries in Python such as OpenCV and
tesseract. By using Pandas, the data obtained was
The text fro m the cropped image is now read using
stored in the form of excel sheet.
tesseract library in Python that perform OCR on the
images and extracts the text fro m them. The string
obtained can be stored in a variable and can be used
later. Fig. 8 shows the detected text from the image. VII. REFERENCES

[1] M . Jasmine Pemeena Priyadarsini, K.M urugesan,

B. Converting to .csv file
The obtained text is stored as a variable. The obtained
text now needs to be documented. For this Pandas
lib rary has been used. Pandas is mainly used for data
Srinivasa Rao Inbathini, A.Jabeena, K.Sai Tej, “Bank
Cheque Authentication using Signature”, International
Journal of Advanced Research in Computer Science and
Software Engineering, Volume 3, Issue 5, M ay 2013.
[2] Rohan Vaidya, Darshan Trivedi, Sagar Satra,

978-1-7281-4889-2/20/$31.00 ©2020 IEEE 96

Authorized licensed use limited to: University of Exeter. Downloaded on June 19,2020 at 13:43:17 UTC from IEEE Xplore. Restrictions apply.
Proceedings of the Fourth International Conference on Computing Methodologies and Communication (ICCMC 2020)
IEEE Xplore Part Number:CFP20K25-ART; ISBN:978-1-7281-4889-2

“Handwritten Character Recognition Using DeepLearning”,

International Conference on Inventive Communication and
Computational Technologies(ICICCT), 2018.

[3] https://techterms.com/definition/ocr
[4] https://pypi.org/project/pytesseract/
[5] https://en.wikipedia.org/wiki/OpenCV

[6] Ankit Arora1, Aakanksha S. Choubey2, “Offline

Signature Verification and Recognition using Neural
Network”, International Journal of Science and Research
(IJSR), India Online ISSN: 2319-7064, Volume 2 Issue 8,
August 2013.

[7]https://github.com/alankarmisra/SwiftSignatureView

[8]https://en.wikipedia.org/wiki/Normalization_(image_pro
cessing)

[9] Dr. Neeraj Bhargava1, Anchal kumawat2, Dr.

Ritu Bhargava3,” Threshold and binarization for document
image analysis using Otsu’s Algorithm”, International
Journal of Computer Trends and Technology (IJCTT) –
volume 17 Number 5 Nov 2014.

[10] Abhishek, Lakshmesha K.N, “Thinning Approach

in Digital Image Processing”, International Journal of
Latest
Trends in Engineering and Technology, Special Issue
SACAIM 2017.

[11] Dipti Verma, Sipi Dubey, “Static Signature

Recognition System for User Authentication Based Two
Level Cog, Hough Transform and Neural Network”,
International Journal of Engineering Sciences & Emerging
Technologies, ISSN: 2231 – 6604, Volume 6 Dec. 2013.

[12] Rahul Dubey, Deeraj A garwal, “Offline Signature

Recognition using Hough Transform and Neural Network’’,
International Journal of Engineering Sciences & Research
Technology, ISSN: 2277-9655, July 2013.

978-1-7281-4889-2/20/$31.00 ©2020 IEEE 97

Authorized licensed use limited to: University of Exeter. Downloaded on June 19,2020 at 13:43:17 UTC from IEEE Xplore. Restrictions apply.