PROJECT SYNOPSIS REPORT

On

INTELLIEXTRACT: AI-BASED VIDEO

TEXT EXTRACTION SYSTEM
Submitted in Partial Fulfillment for the Award of
BACHELOR OF TECHNOLOGY

In

Computer Science & Engineering

(BATCH: 2025)

By

Kartik Gupta (2155029)

Mani Gupta (2155030)

Under the Guidance

Of

Mr. Dileep Kumar Yadav

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERINNG

FACULTY OF ENGINEERING AND TECHNOLOGY
(Uma Nath Singh Institute of Engineering and Technology)

VEER BAHADUR SINGH PURVANCHAL UNIVERSITY,

JAUNPUR (U.P.)
 INTELLIEXTRACT: AI-Based Video Text Extraction System

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING

FACULTY OF ENGINEERING AND TECHNOLOGY
(Uma Nath Singh Institute of Engineering and
Technology)
VEER BAHADUR SINGH PURVANCHAL
UNIVERSITY, JAUNPUR (U.P.)

-------------------------
CERTIFICATE
-------------------------

Certified that the project synopsis entitled “IntelliExtract: AI-Based Video

Text Extraction System” submitted by Kartik Gupta [2155029] and Mani

Gupta [2155030] in the partial fulfillment of the requirements for the

award of the degree of Bachelor of Technology in Computer Science &
Engineering of Veer Bahadur Singh Purvanchal University, Jaunpur
(U.P.) is a record of students’ proposed work carried under my
supervision and guidance. The synopsis report is not been submitted for
the award of any other degree to the candidate.

Mr. Dileep Kumar Yadav

Assistant Professor
(Project Guide)

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page ii

 INTELLIEXTRACT: AI-Based Video Text Extraction System

DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING

FACULTY OF ENGINEERING AND TECHNOLOGY
(Uma Nath Singh Institute of Engineering and
Technology)
VEER BAHADUR SINGH PURVANCHAL
UNIVERSITY, JAUNPUR (U.P.)

-------------------------
DECLARATION
-------------------------

We hereby declare that the project synopsis entitled “IntelliExtract: AI-

Based Video Text Extraction System” submitted by us in the partial
fulfillment of the requirements for the award of the degree of Bachelor of
Technology in Computer Science & Engineering of Veer Bahadur Singh
Purvanchal University, Jaunpur (U.P.), is record of our proposed work
under the supervision and guidance of Mr. Dileep Kumar Yadav
(Assistant Professor).

To the best of our knowledge this project synopsis has not been submitted
to Veer Bahadur Singh Purvanchal University, Jaunpur (U.P.) or any
other University or Institute for the award of any other degree.

Kartik Gupta Mani Gupta

2155029 2155030

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page iii
 INTELLIEXTRACT: AI-Based Video Text Extraction System

ABSTRACT
This paper introduces IntelliExtract, a video text extraction system
designed to accurately capture and extract text from video frames in real
time. Leveraging cutting-edge machine learning algorithms and computer
vision techniques, IntelliExtract is capable of processing diverse video
formats and environments to identify, detect, and extract both printed and
handwritten text from video streams. The system is built using an
intuitive user interface for seamless interaction, allowing users to upload
videos, preview the text extraction process, and retrieve results
efficiently.
Keywords - Key Frame, Frame Selection, Video Indexing, Keyword
Selection, Indexing, Content Retrieval, Text Extraction, Detection,
Binarization, edge, connected component, Frame Extraction, Text
Recognition, Keyword Indexing.

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page iv

 INTELLIEXTRACT: AI-Based Video Text Extraction System

ACKNOWLEDGEMENT
This work is just not an individual contribution till its completion. We
take this opportunity to express a deep gratitude towards our teachers for
providing excellent guidance, encouragement, and inspiration throughout
the training work, without their invaluable guidance this work would
never have been a successful one. We would like to express deepest
appreciation towards our Project Guide Mr. Dileep Kumar Yadav. At last,
we must express our sincere heartfelt gratitude to our HOD Dr. Vikrant
Bhateja, and all the teachers of Computer Science & Engineering
Department, who helped us directly or indirectly during this course of
work.

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page v

 INTELLIEXTRACT: AI-Based Video Text Extraction System

TABLE OF CONTENTS

Certificate ii

Declaration iii

Abstract iv

Acknowledgement v

Table of Contents vi
List of Figures vii
1. Introduction 1-21-2
1.1 Overview 1
1.2 Background 1 1
1.3 Stages Of Text Extraction 2 1
1.4 Project Concept 2 2
2. Review of Related Work 3-53-5
2.1 Methods For Key Frame Selection 3 3
2.2 Methods For Text Extraction 4
2.3 Inferences 5
3. Problem Definition 6 6
3.1 Motivation 6 6
3.2 Aim of the Project 6 6
3.3 Project Objectives 6 6
4. Proposed Design Methodology 7-9
7-9
5. Hardware/Software Requirements & Specifications 10 10
6. Applications of Proposed Project 11-12
11-12
Appendix- ‘A’: List of Abbreviations Used viiiviii
Appendix- ‘B’: List of Common Symbols Used ix
ix
References & Bibliography x
x

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page vi

 INTELLIEXTRACT: AI-Based Video Text Extraction System

LIST OF FIGURES

FIG. NO. FIGURE NAME PAGE NO.

Fig. 2.1 Flowchart of the key frame extraction method 4 4
Fig. 4.1 Flow of Model 88
Fig. 6.1 Opening window of GUI 1111
Fig. 6.2 Select Video in GUI 1111
Fig. 6.3 Indexing of Searched data 1212
Fig. 6.4 Data Not found message 1212

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page vii
 INTELLIEXTRACT: AI-Based Video Text Extraction System

CHAPTER 1
INTRODUCTION
In videos there are different types of text objects. These objects contain information
about videos such as logo of a university which tells university name and various texts
which provide the contents about the video. That’s why extraction of text is important
for video indexing and information retrieval. In this report we have done the exactly
the same thing and returned the text present in the indexed video in the order of their
appearance.

1.1 OVERVIEW
In this project, methods of how to extract proper text from videos are discussed and
also which types of tools are used which method gives how much accuracy shown we
are currently devel- oping tools for indexing video archives for later reuse, a system
for content analysis of videos in which text appearance is different. These all things are
also dependent on their efficient computational support, combining indexed image and
video analysis and processing tools. Now a days in text extraction rapid developments
are shown hundreds of researcher try to do this in proper way and any research paper is
published. Text extraction approaches for videos proposed respectively. In this project,
we mainly concentrate on the approaches proposed for text extraction in videos in the
most recent 5 years and how to get proper text from videos. To summarize and discuss
the recent progress in this research area.

1.2 BACKGROUND
In recent years the availability of videos are growing rapidly over internet specially on
youtube. The text extraction is used for searching important information from video
data sets. Using this extracted text anybody can get an idea about the videos. For
categorizing the extracted text play important role as a key sign. It is also used to
determine the content of the video. Video text extraction is identified as one of the
key components of the video analysis and retrieval system. Video text extraction can
be used in many applications, like multilingual video information access, semantic
video indexing, video security and surveillance etc. In every video which contain text

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page 1

 INTELLIEXTRACT: AI-Based Video Text Extraction System

usually persists for at least some seconds, because of human viewers so that they read
it and understand easily.

1.3 STAGES OF TEXT EXTRACTION

There are different stages of text extraction from videos which are given below-
1. Text detection- In a video frame finding that regions which contain text.
2. Text localization- Combine different text regions into text instances and generating
a set of tight boundary areas around all text instances.
3. Text tracking- Continue to follow a text event as it moves or changes continuously
or not over time and determining the different (temporal and spatial) locations.
4. Text recognition- Performing OCR on the indexed text frame. Occasionally
recognition step is deleted in favour of applying OCR on colour/grey level images.
For extraction of text different techniques are used by many researchers and which
can be classified later. According to different programs and title of that program text
is abundant in videos.

1.4 PROJECT CONCEPT

As from our side we have tried our best to create a system which extracts text from
videos then after it retrieves relevant information from the extracted text. The project
will be completed in three phases-
1. Operation done on video.
2. Text Extraction from Videos.
3. Use Relevant information from Extracted Text.
To document the progress of the system we have created a detailed report and concise
presentation.

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page 2

 INTELLIEXTRACT: AI-Based Video Text Extraction System

CHAPTER 2
REVIEW OF RELATED WORK
2.1 LITERATURE SURVEY
Relevant Information from frames of indexed video is something which has become a
new phenomenon upon which many research papers are being published and still the
searching continues to go on. Although it’s tedious and complex subject but due to its
tremendous use it’s a hot potato for many years. The research papers which has been
published regarding the same is thoroughly analysed and referred for further
understanding. The techniques which are mentioned the papers are explained in
subsequent parts of the project research. As we move in ahead we discuss different
phases of project.

2.2 METHODS FOR KEY FRAME SELECTION

2.2.1 Key Frame for Video Copyright Protection
There are some distinct features about the key frame for video copyright protection.
So,the key frame for video copyright protection is defined firstly before video pre-
processing and key frame extracting. The key frames should meet the following three
conditions:
1. The key frame is within a certain range to allow viewers to have subjective
perception about the video content. Images with low gray value in Fig.2.1 are
extracted from a single video, which is difficult for almost viewers to recognise the
content.
2. The final key frame sequence must be arranged in chronological order consistent
with original video sequence, in order to satisfy text extraction features and to be
different from the short promotion trailer.
3. Appropriate redundancy of some key frames is allowed to ensure the periods or
intervals along the processing of video content. Many Images in a video, which are
with similar content, that is to say, one judge in the show every once in a while.

2.2.2 Two-Stage Method for Key Frame Extraction

In a key frame extraction for digital video copyright protection. First, a digital video is
de- composed into video frames. The downloaded video from the network includes

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page 3

 INTELLIEXTRACT: AI-Based Video Text Extraction System

several video formats, such as f4v, flv and mp4. In order to improve the universality
of video key extraction algorithm, the present method does not consider the specific
format and video stream structure, and the video is decoded before the processed
video frame decomposition. It is seen that the program to extract key frame is divided
into two steps.

2.2.3 Performance Analysis of Key Frame Extraction Methods

A key frame extraction method based on frame difference with low level features is
proposed for video copyright protection. Exactly, a two-stage method is used to
extract accurate key frames to cover the content for the whole video sequence. Firstly,
an alternative sequence is obtained based on color characteristic difference between
adjacent frames from original sequence.

Figure 2.1

Secondly, the final key frame sequence is obtained by analyzing structural

characteristic difference between adjacent frames from the alternative sequence. Two
stage method is used mostly because of frame difference value. This method calculate
frame difference value is more accurate than video copyright method.

2.3 METHODS FOR TEXT EXTRACTION

2.3.1 Region Based Approach
In recent years there is huge increase in multimedia libraries. The size of multimedia
data is growing exponentially. Main reason for growing multimedia data is increasing
in numbers of television channels that are broadcasting every day. Also due to
advancement in technology cameras became affordable, memory device is
inexpensive, multimedia data is increasing every second. Surveillance cameras to
broadcast videos from phone’s camera and various social net- working application are
adding enormous multimedia data.

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page 4

 INTELLIEXTRACT: AI-Based Video Text Extraction System

2.3.2 Texture Based Approach

Texture based technique use the assumption that text in indexed frames carries
distinct textural properties, which may be used to differentiate it from the background.
Generally to extract the textural properties of a text region in an image. The usual
approach is to use a classifier trained to divide regions to textual/non-textual based on
texture features. These methods use machine learning and are less heuristic-based, but
they are more computational expensive.

2.3.3 Edge Based Approach

Text embedded in document in complex coloured and textured backgrounds are
increasingly common today, for example, in web pages, in magazines and
advertisements. Efficiently detection and extracting of text from these documents is a
challenging problem. The procedure generated for ordinary documents, such as
binarization by adaptive thresholding are not applicable in general, because it is
almost impossible to find an optimal threshold or thresholds to preserve meaningful
information and to discard unnecessary one.

2.4 INFERENCES
Building the INTELLIEXTRACT model requires thoughtful selection of tools (like
OpenCV and Tesseract), effective preprocessing (e.g., adjusting contrast in video
frames), and a robust model architecture such as combining EAST for text detection
and CRNN for recognition. Key considerations include handling different text
orientations, optimizing processing speed by filtering frames without text, and
training on diverse datasets for fonts and languages. To achieve accuracy and
efficiency, advanced preprocessing techniques and custom datasets are essential,
particularly for domain-specific needs. By leveraging batch processing and cloud
resources, the model can be scaled for large video datasets, making IntelliExtract
adaptable for real-world applications.

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page 5

 INTELLIEXTRACT: AI-Based Video Text Extraction System

CHAPTER 3
PROBLEM DEFINITION
3.1 MOTIVATION
There are different types of methods to extract the text from videos. These methods
are for specific applications including page segmentation, license plate location and
content-based video indexing. After studying such types of text extraction method it is
not easy task to design a general text information extraction (TIE) system. In videos
there are different types of variations such as complexity of background, font size,
color, style, alignment, brightness that’s why design of a TIE system is tough. These
variations play a important role to not working properly a automatic TIE system.
After studying different methods of text extraction analyzing their evaluation results
performance evaluation approaches not only search for answers to many questions
such as: Which text extraction method is better? Why does performance of different
methods is varying in different types of dataset ? Which types of error comes at the
time of indexing ? These questions actually help to develop new ideas to improve the
extraction technology and specific algorithms.

3.2 AIM OF THE PROJECT

The aim of IntelliExtract is to develop an AI-based system capable of accurately
extracting text from video frames in real time. By utilizing advanced machine
learning and computer vision techniques, the project seeks to automate the detection
and retrieval of both printed and handwritten text from various video formats. The
goal is to create a robust solution that simplifies text extraction from dynamic video
content, making it highly useful across fields like education, media analysis, legal
documentation, and archival research.

3.3 PROJECT OBJECTIVES

In this project, methods of how to extract proper text from videos are discussed and
also which types of tools are used which method gives how much accuracy shown we
are currently developing tools for indexing video archives for later reuse, a system for
content analysis of videos in which text appearance is different.

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page 6

 INTELLIEXTRACT: AI-Based Video Text Extraction System

CHAPTER 4
PROPOSED DESIGN METHODOLOGY
The main goal of this methodology is to approach for automated video indexing and
video search from video lecture archives. The methodology further aims to apply
automatic video segmentation and key-frame detection to offer a visual guideline for
the video content extraction in the order of their appearance in the video. Extract
textual metadata by applying video Optical Character Recognition (OCR) technology
on key-frames.

4.1 PROPOSED MODEL

In recent chapter there are some methods discussed which is used for text extraction
from a video and indexing of that content. Video is a collection of different images.
Text extraction from a video is not a easy task because in a video many types of data.
Suppose if there is video where teacher teaches to students using projector then there
are different slides showing on projector. For indexing of content which is present on
slides in video. So for indexing the video content and respective time of user wanted
data we follow the flow.

Figure 4.1

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page 7

 INTELLIEXTRACT: AI-Based Video Text Extraction System

4.2 FRAME GENERATING FROM VIDEO

First we take a video as a input and generate frame using opencv with fps value=30.
Frames as a part of video at a particular instance Even for a small video many frames
are generated.
Number of frames in a video = Time duration of video * 30
Process of Frame Generation:
1. Open the Video file or camera using cv2.VideoCapture()
2. Read frame by frame
3. Save each frame using cv2.imwrite()
4. Release the Video Capture and destroy all windows

4.3 SUPPORT VECTOR MACHINE (SVM)

Support Vector Machines (SVM) are supervised machine learning models widely
used for classification and regression tasks. SVM works by finding a hyperplane (or
decision boundary) in a multi-dimensional space that best separates different classes.
The main goal of SVM is to maximize the margin, which is the distance between the
hyperplane and the closest data points from each class, called support vectors. This
maximized margin helps the model generalize well to unseen data, making it robust
for classification tasks.

In cases where classes are not linearly separable, SVM can use a kernel trick to
transform the data into a higher-dimensional space, where it becomes easier to draw a
separating hyperplane. Common kernel functions include linear, polynomial, and
radial basis function (RBF). SVM is known for its effectiveness in high-dimensional
spaces and its ability to work well even with a limited number of samples, making it
ideal for applications like image recognition, text classification, and bioinformatics.

4.4 OPTIMAL CHARACTER RECOGNITION (OCR)

Optical Character Recognition (OCR) is a technology that converts different types of
documents, such as scanned paper documents, PDF files, or images captured by a
camera, into editable and searchable digital text. OCR works by analyzing the shapes

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page 8

 INTELLIEXTRACT: AI-Based Video Text Extraction System

of individual characters within an image and recognizing them as letters, numbers, or

symbols.
4.4.1 Working of OCR
• Image as a input has to be given from which we have to extract text. This
image is either to be a picture or scanned one.it is stored as bitmap format.
• To make the image properly aligned we need to apply de-skewing i.e tilting in
clockwise and anti-clockwise direction, and also remove noise to improve the
quality of image.
• Binarization is done to convert an image to black and white. It is used to
separate the text from the background. This is curious because inaccurate
binarization will cause lot of issues.
• Detection and removal of lines.
• Combined and broken character analysis.
• Isolation of characters, multiple characters that are connected must be
separated and single characters that are into multiple pieces must be
connected.
• Classification of characters, the text are divided into lines and then into
characters and after that character is recognized using various algorithm.
algorithm such as Matrix matching and feature extraction is used to produce a
ranked list of candidate character.
• Dictionary support, It helps to improve the recognition quality. characters like
"C" and "G" can look similar, so dictionary can help to make decisions.
• At last the result is saved in the selected output format such as PDF, DOC etc.

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page 9

 INTELLIEXTRACT: AI-Based Video Text Extraction System

CHAPTER 5
HARDWARE/SOFTWARE REQUIREMENTS &
SPECIFICATIONS
For developing IntelliExtract: AI-Based Video Text Extraction System, the following
hardware and software are recommended to ensure smooth development & operation.

5.1 HARDWARE REQUIREMENTS

1. Computer/Laptop:
• Processor: Intel i5 or AMD Ryzen 5 and above
• RAM: 8 GB (minimum), 16 GB (recommended for better performance)
• Storage: 256 GB SSD (minimum), 512 GB SSD or more (recommended)
• Graphics: Integrated graphics (for basic testing) or dedicated GPU.
2. Internet Connection:
• Stable connection for downloading software , and testing cloud-based
functionalities.

5.2 SOFTWARE REQUIREMENTS

1. Operating System:
• Windows 10/11, macOS, or a Linux-based OS (Ubuntu recommended).
2. Programming Languages:
• Python: For implementing the AI and machine learning models, especially for
text detection and recognition.
3. Development Tools/Frameworks:
• OpenCV: A popular library for image and video processing, used for handling
video frames and detecting text regions.
4. IDEs/Text Editors:
• Jupyter Notebook or PyCharm: For developing and testing the AI models in
Python.
5. APIs/Libraries:
• Tesseract OCR: An optical character recognition (OCR) engine for
recognizing and extracting text from video frames.

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page 10

 INTELLIEXTRACT: AI-Based Video Text Extraction System

CHAPTER 6
APPLICATION OF PROPOSED PROJECT
We design this model in frontend and backend. In frontend we create a GUI where
title bar is there, a canvas window where video is running continuously. Right side of
this canvas window result window is there where accuracy of model for that particular
text and indexing of text is there. At the below of that window video controller option
is there like start button, stop button, restart video button, volume controller, find
button. In backend all other process like frame generation, key frame selection, video
frame indexing, text extraction process is executed and show result on our GUI
window Please refer Figure 6.1 and Figure 6.2.

Figure 6.1

Figure 6.2

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page 11

 INTELLIEXTRACT: AI-Based Video Text Extraction System

6.1 VIDEO FRAME INDEXING AND APPLY TEXT

EXTRACTION
For video frame indexing we save time of each frame. User wants which data to find,
we extract data from each key frame and compare with user data. If data is found in
any frame then respective time and accuracy shows else a message shown. Please
refer the below figure (Fig. 6.3 and Fig. 6.4) for better understanding.

Figure 6.3

Figure 6.4

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page 12

 INTELLIEXTRACT: AI-Based Video Text Extraction System

APPENDIX- ‘A’: LIST OF ABBREVIATIONS

USED

1. CC Connected Component
2. DR Detection Rate
3. ECR Edge Change Ratio
4. FAR False Alarm Rate
5. FD Frame Difference
6. OCR Optimal Character Recognition
7. PDE Partial Differential Equation
8. PR Precision Rate
9. RR Recall Rate
10. SSD Sum of Squared Difference
11. SVM Support Vector Machine
12. TIE Text Information Extraction

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page viii
 INTELLIEXTRACT: AI-Based Video Text Extraction System

APPENDIX- ‘B’: LIST OF COMMON SYMBOLS

USED

1. ⊕ (Plus or Add): Used to represent the addition of data or elements, possibly

for concatenating text or processing multiple video frames.
2. → (Arrow): Represents the flow of data or information, such as moving from
video frames to text detection.
3. ⊗ (Multiplication or Tensor Product): Used in machine learning
algorithms, particularly in matrix operations during the training and
application of neural networks.
4. Σ (Sigma): Summation symbol, often used in algorithms like backpropagation
in neural networks to calculate error gradients.

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page ix

 INTELLIEXTRACT: AI-Based Video Text Extraction System

REFERENCES & BIBLIOGRAPHY

[1] Gongqing, W., Jun, H., Li, L.L., et al.: Online content extraction based on
label path feature fusion. J. Softw. 27(3), 714–735 (2018).
[2] Wu, Jung G.Q., Hu, J., Li, L., Xu, Z.H., Liu, P.C., Hu, X.G., Wu, X.D. :Web
news extrac- tion via tag path feature fusion. Ruan Jian XueBao/J.Softw.
27(3), 714–735 (2018).
[3] Jiazhen, C., Yan, G., Qiang, L., et al.: An automatic text extraction method for
short text web pages. Chin. J. Inf. Sci. 30(1), 8–15 (2016).
[4] Q. Ye, D. S. Doermann, “Text Detection and Recognition in Imagery: A
Survey”, IEEE Transactions on Pattern Analysis and Machine Intelligence,
Vol. 37(7), pp. 1480-1500, 2015.
[5] V. Khare, P. Shivakumara, P. Raveendran, M. Blumenstein, “A blind
deconvolution model for scene text detection and recognition in video”,
Pattern Recognition, Vol. 54, pp.128- 148, 2016.
[6] A. Gonzalez, L. M. Bergasa, J. J. Yebes. "Text detection and recognition on
traffic pan- els from street-level imagery using visual appearance", IEEE
Transactions on Intelligent Transportation Systems, Vol. 16(3), pp. 228-238,
2015.
[7] A. K. Bhunia, A. Das, P. P. Roy, U. Pal, “A Comparative Study of Features of
Handwrit- ten Bangla Text Recognition”, In Proceedings of International
Conference on Document Analysis and Recognition, pp.636-640, 2015.
[8] Zhong, A., X. Peng, X. Zhuang, P. Natarajan, H. Cao,Ohya “Text detection
and recognition in natural scenes and consumer videos”. In Proceedings of
International Conference on Acoustics, Speech and Signal Processing, pp.
1245-1249, 2014.
[9] H. Yang, B. Quehl, H. Sack, “A framework for improved video text detection
and recogni- tion”, Multimedia Tools and Applications, Vol. 69(1), pp.217-
245, 2014.

CSE, UNSIET, Veer Bahadur Singh Purvanchal University, Jaunpur Page x