International Journal of Science and Research (IJSR)

ISSN (Online): 2319-7064

Index Copernicus Value (2016): 79.57 | Impact Factor (2015): 6.391

Photo Restoration using Open CV in Python: A

Computer Vision Based Project
Subhojit Paul1, Saikat Ghosh2, Dhrupad Saha3, Sourajit Paul4, Rajesh Dusad5, Mrinmoy Biswas6
1, 2, 3, 4, 5, 6
Department of Electrical Engineering, UEM Kolkata, Kolkata, India

Abstract: At present image processing has become one of the very famous subjects for research area among rapidly growing
technologies within engineering. This paper will explain how a damaged or corrupted image or photograph (cracked, torn or folded) can
be restored using OpenCV(Open Source Computer Vision) in Python. Here we feed a damaged or corrupted image in the program along
with a masked image as input and eventually we will get restored image as output. There are some methods like Denoising, Gray scaling,
Thresholding, Dilation etc and few inbuilt function and algorithms are used to construct the code. Thus this is a very interesting, simple,
and fascinating image processing technique and can be used for several purposes like restoring old, damaged or corrupted image for
archiving purposes.

1. Introduction We should counter one problem with photographs when

they’ve been used in displays or albums; they are often put
Photos are generally collection of memories that might be onto paper or cardboard, which degrades at a different rate to
very special for an individual. Thus if any kind of damage the photograph. They absorb moisture differently and are
occurs, we should be able to recover or restore the original prone to insect damage. Our main objective is to minimise
photo. In a broader approach, the photographs we usually see these effects, clean and stabilise the photograph again
which have been preserved and restored from important
moments in history. Hindsight's a wonderful thing, and we 2. System Design
often don't understand that, when we are capturing
something that generations will be interested in years from The photo-restoration process is executed with the help of
now. You can't keep everything, but when you think two different software -
carefully about which photos to delete and which to keep. 1) OpenCV (Open Source Computer Vision)
This is really based on luck. Photographs are one of the best 2) Python
ways of recording what. happens, they can capture a place,
events in a place, or people's lives. Photographs are OpenCV(Open Source Computer Vision):
something that people have used more than 100 years to OpenCV is one type of library of programming functions. It
record, so they are pretty useful when we looked back and is mainly implemented at real-time computer vision
notice how life has changed. People are more keen to look at technology. Initially Intel introduce this kind of technology
the picture than to read the whole descriptive lines. and later it was supported by Willow Garage. Nowadays it
is maintained by Itseez. The library is cross-platform which
We should be much selective in case of capturing a single means that two different kinds of platforms can participate
moment and to showcase it. For an example-You don't need separately but work as an integrated system. OpenCV free
to capture 20 photos for showing the same thing, so just keep for use under the open-source BSD license. As OpenCV
the best ones unless they became the part of a series basically using two separate platforms hence it is refer as
chronicling a time or event. In recent days the popularity of multi-platform or platform-independent software. OpenCV is
digital cameras means that there are a lot more photographs represented in C language/C++.
being taken and almost everyone has a camera on their
mobile phone. As a result about any event is being Deep learning frameworks like TensorFlow,
photographed and documented by several people. In the past, Torch/PyTorch and Caffe, is explained by OpenCV.
it was difficult to find photographs of events but now there’s Primary interface of OpenCV is C++, but still it is also
almost too much being produced and you are filled with comfortable C language also. There are several segments in-
information. 1)Python, 2) Java and 3)MATLAB/OCTAVE. Whenever
the library discover Intel’s Integrated Performance Primitives
The damaged photo doesn’t mean that we should discard on the system, it will be use these proprietary optimized
them. If photographs are damaged you don't necessarily have routines by means of accelerate itself. This is able to run on
to write them off. There is a lot we can do digitally ourselves following desktop OS: Windows, macOS, Linux, FreeBSD,
these days to bring them back to an acceptable condition. NetBSD, OpenBSD. Eventually OpenCV mobile OS:
You can also send off a hard copy to be restored Android, iOS, Maemo, BlackBerry 10.
professionally if it’s a photograph that's really important to
you.

Thus we are introducing a programmable language using

Open CV and Python to restore the damaged images in
proper way.

Volume 7 Issue 1, January 2018

www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
Paper ID: ART20179794 DOI: 10.21275/ART20179794 1967
 International Journal of Science and Research (IJSR)
ISSN (Online): 2319-7064
Index Copernicus Value (2016): 79.57 | Impact Factor (2015): 6.391
Now, this masked file is thresholded and converted into a
binary image. Thresholding is an effective way which
enables us for segmentation of the image. It partitions an
image by isolating into foregrounds and backgrounds, i.e., it
helps in detecting the marked areas that needs to be repaired

After the above step, Dilation is incorporated. As

Thresholding reduced the thickness of the marked areas
Figure 1: Logo of OpenCV which is not desired, dilation is necessary. Dilation enables
us to add pixels over a region by defining a kernel value
Python leading in making the marked lines thicker. Then this dilated
Python is an interpretable high-level language. However, it is masked file is saved into our machine.
the first general utility programming language. It is
developed by Guido van Rossum and was first released in And here comes the core command which basically repairs
1991. Python has a design motive that encourages code the damaged areas: Image Inpainting Function. This
clarity, and a syntax that doesn’t needs long blocks of technique works upon the Fast Marching Method. The
functions. algorithm starts from the boundary of the marked area and
goes on filling the areas with an average value of the pixels
Python interpreters are available for most of the operating from a small circular neighbourhood of that area. We have
systems. CPython, is an open source software. The software used TELEA method here which is better as it integrates
has a community-based development model. CPython is more seamlessly into the image. Finally the image is ready,
handled by the non-profit Python Software Foundation. rendered as well as restored.

4. Flowchart

Figure 2: Logo of Python

It aims for a dynamic type system and automatic memory

management. It is compatible for multiple paradigms, which
includes object-oriented, functional, procedural and
imperative, and has a vast and comprehensive standard
library.

Anaconda (Python distribution):

Anaconda is an open-source distribution of Python. The R
programming languages is using for large-scale data
processing, predictive analytics and scientific computing,
that aims to simplify package management and development.
Package versions are managed more easily by package
management system conda.

Figure 3: Logo of Anaconda

3. Methodology
Initially the path of the damaged file is mentioned and is
loaded in our program. To enhance the final output image,
the damaged photo is processed by Denoising it by means of Figure 4: Flowchart of the program
Fast Means Denoising method. This reduces the coarseness
of the image and makes the image smooth. 5. Result
Next, a greyscale masked image of the same damaged image In this section we will get to know how the damaged or
is created by marking the regions which are to be repaired corrupted image is being developed to restore image in step
and is also loaded. The marking is done by creating strokes by step format as defined in the flowchart section. The result
over the damaged area with a brush of white colour edited in is as following:
MS Paint for Windows or any similar software in Mac or
Linux Ubuntu.

Volume 7 Issue 1, January 2018

www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
Paper ID: ART20179794 DOI: 10.21275/ART20179794 1968
 International Journal of Science and Research (IJSR)
ISSN (Online): 2319-7064
Index Copernicus Value (2016): 79.57 | Impact Factor (2015): 6.391

Figure 5: Original Damaged image

Figure 8: Grayscaled Mask image

Figure 6: Denoised image

Figure 9: Threshold Binarization

Figure 7: Mask image

Figure 10: Threshold Binarized image after Dilation

Volume 7 Issue 1, January 2018

www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
Paper ID: ART20179794 DOI: 10.21275/ART20179794 1969
 International Journal of Science and Research (IJSR)
ISSN (Online): 2319-7064
Index Copernicus Value (2016): 79.57 | Impact Factor (2015): 6.391
International Conference on Electrical Engineering,
Computer Science and Informatics (EECSI).
[8] Hao Jiang, Thomas Gonnot, Won-Jae Yi and Jafar
Saniie, “Computer Vision and Text Recognition for
Assisting Visually Impaired People using Android
Smartphone “,2017 IEEE International Conference on
Electro Information Technology (EIT).
[9] Aniket V. Patil, Mrinai M. Dhanvijay, “Engraved
Character Recognition Using Computer Vision To
Recognize Engine And Chassis Numbers”, 2015
International Conference on Information Processing
(ICIP).
[10] T. Arrighi , J. E. Rojas , J.C. Soto , C. A. Madrigal , J.
A. Londoño, “Recognition and Classification of
Numerical Labels Using Digital Image Processing
Techniques “,2012 XVII Symposium of Image, Signal
Processing, and Artificial Vision (STSIVA).
Figure 11: Restored image of damaged image

6. Conclusion
In this paper we discussed about Image restoration by the
help of Inpainting function using the language of Python
with the help of OpenCV software. This program gives us
facility to restore image from damaged condition for
archiving purpose of any old, torn photograph etc which can
be considered as application of this program. We have
successfully done the restoration of some photographs in our
project, so we conclude that we can easily restore all types of
damaged photograph using this effective and efficient
program code.

References
[1] Nidhi, “Image Processing and Object Detection”,
International Journal of Applied Research 2015;1(9):
396-399
[2] Kruti Goyal, Kartikey Agarwal, Rishi Kumar , “Face
Detection and Tracking”, International Conference on
Electronics, Communication and Aerospace Technology
ICECA 2017
[3] Y ue Yaru, Zhu Jialin, “Algorithm of Fingerprint
Extraction and Implementation Based on OpenCV”,
2017 2nd International Conference on Image, Vision and
Computing,978-1-5090-6238-6/1 7©20 17 IEEE
[4] Chaoxing Huang ,Dan Chen ,Xusheng
Tang,”Implementation of Workpiece Recognition and
Location Based on Opencv”, 2015 8th International
Symposium on Computational Intelligence and Design.
[5] M. Ashok Kumar, R.R. Tewari, “Human gestures and
recognizing faces in natural expressions by considering
computer vision techniques on mobile devices”,
Intelligent Computing and Control Systems (ICICCS),
2017 International Conference.
[6] Cristhian Rosales, Luis Jácome, Jorge Carrión, Carlos
Jaramillo, Mario Palma, “Computer Vision for detection
of body expressions of children with cerebral palsy”,
2017 IEEE Second Ecuador Technical Chapters Meeting
(ETCM).
[7] G.D. Illeperuma, D.U.J. Sonnadara, “Computer Vision
Based Object Tracking as a Teaching Aid for High
School Physics Experiments”, Proc. EECSI 2017,
Yogyakarta, Indonesia, 19-21 September 2017, 2017 4th
Volume 7 Issue 1, January 2018
www.ijsr.net
Licensed Under Creative Commons Attribution CC BY
Paper ID: ART20179794 DOI: 10.21275/ART20179794 1970