Development and Analysis of Enhanced Image Inpainting Approach

International Journal of Trend in Scientific Research and Development (IJTSRD)
Volume 6 Issue 4, May-June 2022 Available Online: www.ijtsrd.com e-ISSN: 2456 – 6470
@ IJTSRD | Unique Paper ID – IJTSRD50366 | Volume – 6 | Issue – 4 | May-June 2022 Page 1731
Development and Analysis of
Enhanced Image Inpainting Approach
Yogita1
, Shalini Bhadola2
, Kirti Bhatia2
, Rohini Sharma3
*
1
Student, Sat Kabir Institute of Technology and Management, Bahadurgarh, Haryana, India
2
Assistant Professor, Sat Kabir Institute of Technology and Management, Bahadurgarh, Haryana, India
3
Assistant Professor, GPGCW, Rohtak, Haryana, India
ABSTRACT
Inpainting is a restoration technique that involves filling in damaged,
degraded, or missing areas of artwork to create a full image. Oil or
acrylic paints, biochemical photography prints, sculptors, and digital
photos and video are all examples of physical and digital art mediums
that can be used in this process. We have developed a model which
in-paint a corrupted image using three different sparse representation-
based approaches. We have used K-SVD (Singular Value
Decomposition, ORTHOGONAL MATCHING PURSUIT (OMP),
and Delaunay Triangulation based Interpolation. It has also become
apparent from the results that inpainting on natural images appears
decent when not requiring too big patch size. As the patch sizes
increase, to be able to cover large masked areas, the reconstruction
will be smoother. There might be useful at times to have an algorithm
like this that could be used for smoothing and inpainting
simultaneously. However, if the contrast of the image is to be
unharmed some other method should be considered. It is therefore
concluded that even though sparse reconstructive methods appear
impressive at first glance they are lacking when it comes to using
large image patches, which is required when it comes to inpainting
phase maps.
KEYWORDS: Image Inpainting, Sparse Representation
How to cite this paper: Yogita | Shalini
Bhadola | Kirti Bhatia | Rohini Sharma
"Development and Analysis of
Enhanced Image Inpainting Approach"
Published in
International Journal
of Trend in
Scientific Research
and Development
(ijtsrd), ISSN: 2456-
6470, Volume-6 |
Issue-4, June 2022,
pp.1731-1737, URL:
www.ijtsrd.com/papers/ijtsrd50366.pdf
Copyright © 2022 by author(s) and
International Journal of Trend in
Scientific Research and Development
Journal. This is an
Open Access article
distributed under the
terms of the Creative Commons
Attribution License (CC BY 4.0)
(http://creativecommons.org/licenses/by/4.0)
INTRODUCTION
The technique of calculating the clear, original image
from a damaged image is termed image regeneration.
Corruption can be seen in the form of motion blur,
noise, and misfocus of camera. Image regeneration is
achieved by correcting the blurring phenomenon,
which is conducted by imaging a point source and
recovering the visual features missed during the
blurring process using the point source picture,
commonly known as the Point Spread Function.
Image enhancement differs from image restoration in
that the second is intended to highlight characteristics
of the image that make it more attractive to the
spectator, rather than inevitably producing true data
from a scientific standpoint. Image processing
algorithms given by imaging packages do not use an a
priori description of the method that formed the
image. Noise can be efficiently eliminated with
picture enhancement bysurrendering some resolution,
but this is not adequate in various situations.
Resolution in the z-direction of a fluorescent
microscope is already poor. To retrieve the item,
more sophisticated methodologies must be used. A
restoration process called inpainting includes filling
in damaged, deteriorated, or missing sections of
artwork to complete the picture [10]. The approaches
used in inpainting are determined by the required
outcome and the sort of image being processed.
Physical and digital art have completely different
approaches to filling up the voids.
Related Work
Image processing is a very popular field of research.
A plenty of research has been accomplished in
different aspects of this area, such as recognition of
face from a collection of faces [1], sentiments
identification[2-3] removal of noise form the
picture[4-8]. Many tools are now available that can
restore lost or broken parts of digital pictures and
films. Adobe Photoshop is the most well-known
software for working with digital photos. Because
IJTSRD50366

International Journal of Trend in Scientific Research and Development @ www.ijtsrd.com eISSN: 2456-6470
digital files may be replicated, any necessary changes
should be performed to the replica file, while the
initial files should be archived. Inpainting has become
an involuntary procedure that can be carried out on
digital photographs, thanks to the varied capabilities
of the digital camera and the digitization of historical
photos. Inpainting techniques can be used for more
than only scratch removal; they can also be used for
object expulsion, word withdrawal, and other
automated picture and video alterations. In addition,
they can be seen in picture compression and super
resolution applications. It is used to reverse, repair, or
alleviate deterioration in film in cinematography and
filmmaking. It can also be employed to remove
damage signs, the obsolete date from photos, and
items for artistic purposes. This method can be used
to substitute any missing blocks in picture encoding
and communication, such as in a video streaming. It's
also useful for removing branding from videos.
Inpainting based on deep learning neural networks
can be employed to decensor images. In the literature,
there are three primary classes of 2D picture
inpainting methods. The first is structural (or
geometric) inpainting, followed by texture inpainting,
and finally a mixture of the two. All of these
inpainting techniques are similar: they fill in the gaps
using information from recognized or correct image
portions, analogous to how real photographs are
repaired. The practice of inpainting has its origins in
the renovation of painted images. "The phrase
inpainting pertains to the reparation of paint losses—
focusing at the recompositing of the missing
components of a picture in terms of improving its
viewpoint by creating reparations less noticeable [11].
Also, inpainting tries to enhance the overall
appearance of artwork by replacing missing or
spoiled areas with systems and materials that are
comparable to the original artist's work. It is critical
to retain complete records of the primary condition of
the photos, treatments performed and justifications
for treatments, as well as original copies when
applicable, including all applications of inpainting.
ELEMENTARY PROCEDURE OF IMAGE INPAINTING
The process of region filling in digital photographs after information loss is an important part of image
processing. Image inpainting pertains to rebuilding techniques that are used to eliminate damaged or undesired
objects from a picture in such a genuine way that an indistinct spectator would not detect any differences and
mistake the outcome for the original. Structural inpainting techniques, textural inpainting methods, and hybrid
approaches are the three basic categories of restoration procedures. Regardless of these 3 groups, methods can be
classified as PDE-centered methods, semiautomatic inpainting procedures, surface amalgamation approaches,
procedures based on prototypes, and amalgam systems. In this paper, we have used three approaches of image
inpainting (K-SVD, OMP and Interpolation) and have compared their results. The complete process of image
inpainting has ben revealed in figure 1.
Fig 1: A Complete Inpainting Procedure
SPARSE RECONSTRUCTION
Sparse reconstruction is a series of strategies for reconstructing MR pictures from significantly under-sampled k-
space data using image attributes that are known a priori. Several applications necessitate the recovery of image
missing portions. For example, image and video communications across error-prone networks utilising block-
based coders may result in block losses. In an image processing system, flaws in the capture, storage, or other
processes cause mistakes, necessitating the usage of restoration methods to estimate missing portions. Without
any mistake repair data transmitted by the encoder, decoder side recovery algorithms function on the received
data, fig 2.

Fig 2: Sparse Method based Image Inpainting
K-SVD (Singular Value Decomposition)
K-SVD is a dictionary learning method that uses a singular value decomposition technique to produce a
dictionary for sparse representations. Iteratively switching between sparse coding the input data using the
existing dictionary and changing the dictionary's atoms to suit better the data., K-SVD is a generalisation of the
k-means clustering approach. The expectation-maximization (EM) algorithm is structurally related to it [12-13].
As follows, K-SVD is a generalization of K-means. K-means clustering can also be thought of as a sparse
representation method. That is, a nearest neighbor finds the best potential codebook to describe the data samples.
The K-means algorithm begins by making an educated estimate about certain items that best define the data. K-
SVD imitates this step by asking for an initial guess of the dictionary, meaning that the starting point is any
matrix such that,
Fig 3, [14] shows the working of K-SVD algorithm and Fig 4 shows the use of K-SVD in our proposed
method.
Fig 3: Workflow of K-SVD algorithm

ORTHOGONAL MATCHING PURSUIT (OMP)
K-SVD relies on a pursuit algorithm. There are several examples of such algorithms. One of the most commonly
used pursuit algorithms is called Orthogonal Matching Pursuit [15].
INTERPOLATION
Interpolation is a technique for predicting the values of pixels at unknown positions using pixels with known
values. The division of the image into overlapping patches and treatment of each patch based on the natural
image patches is a standard method for achieving interpolation.
Fig 4: Use of K-SVD in Inpainting
PROPOSED METHOD
Step 1: Get the dimensions of the image.
Step 2: Calculate block size of the image. Provide a number of atoms in the dictionary and of pixels
between consecutive patches.
Step 3: Take a picture with missing components.
Step 4: Apply any of the method (Interpolation, OMP or K-SVD) for inpainting of the image.
Interpolation: Use Delaunay triangulation.
OMP: Compute the mask and extract the patches of the image. Extract the noisy image patch.
K-SVD: Extract the patches of the image and compute its mask. Apply discrete transformations.
Step 5: Recover the Image.
Step 6: Compute PSNR (Peak Signal to Noise ration).
A novel approach (sparse-based image inpainting methodology) is used to determine if the image can be
reconstructed properly or not. The adaptive sparse presentation appears to be better than other image restoration
methods. In comparison to the traditional image reconstruction, a minor alteration is made. This article uses
three techniques for image inpainting and then compares the results of these methods
RESULTS
The implementation of the proposed method has several parts. We have provided an image with missing
components (See figure 5). Then three different techniques, Interpolation, OMP and K-SVD are applied
separately to recover the image.

Fig 5: Image with Missing Components
In painted Image after applying K-SVD method (figure 6).
Fig 6: Restored image after applying K-SVD
Image restoration after applying Interpolation (fig 7).
Fig 7: Image restoration after applying Interpolation

Fig 8: Image restoration after applying ORTHOGONAL MATCHING PURSUIT
From the results, it has been observed that K-SVD and OMP techniques are better than the Interpolation method.
Conclusion
We suggested a novel method that uses a sparse
representation method for image inpainting. This
approach computes the dimensions and bloc size of
the image. It creates a dictionary of atoms, and
computes pixels between consecutive patches.
Afterward, it computes the mask and extracts the
patches of the image. Then, it extracts the noisy
image patch. It recovers the missing parts of the
image. Through the outcomes, it is evident that K-
SVD algorithm is better than other two. Although
deep learning is quickly evolving, it is not always a
viable solution to inpainting challenge. The
fundamental reason for this is the absence of image
pairings for training in real-world inpainting
operations. All existing inpainting methods, to our
awareness, are trained on replicated noisy data
acquired by toting AWGN to spotless photographs.
Nonetheless, we discovered that CNNs trained on
simulated data are ineffective for the inpainting
operation in the actual world.
REFERENCES
[1] Sanju, K. Bhatia, Rohini Sharma, Pca and
Eigen Face Based Face Recognition Method,
Journal of Emerging Technologies and
Innovative Research, June 2018, Volume 5,
Issue 6, pp. 491-496.
[2] Ankit Jain, Kirti Bhatia , Rohini Sharma,
Shalini Bhadola, An Overview on Facial
Expression Perception Mechanisms, SSRG
International Journal of Computer Science and
Engineering, Volume 6 Issue 4 - April 2019,
pp. 19-24.
[3] Ankit Jain, Kirti Bhatia , Rohini Sharma,
Shalini Bhadola, An emotion recognition
framework through local binary patterns,
Journal of Emerging Technologies and
Innovative Research, Vol -6, Issue-5, May
2019.
[4] Deepak Dahiya, Kirti Bhatia, Rohini Sharma,
Shalini Bhadola, Development and Analysis of
Enhanced Window Median Filter Approach of
Image Denoising, International Journal Of
Multidisciplinary Research In Science,
Engineering and Technology, Volume 4, Issue
8, August 2021.
[5] Deepak Kumar Shrivastava, Kirti Bhatia,
Shivkant, Rohini Sharma Development and
analysis of mean shift based Video object
tracking tool, International journal of
Innovative Research in computer and
communication engineering, Vol-08, Issue-07,
july 2020.
[6] Deepak Dahiya, Kirti Bhatia, Rohini Sharma,
Shalini Bhadola, A Deep Overview on Image
Denoising Approaches, International Journal of
Innovative Research in Computer and
Communication Engineering, Volume 9, Issue
7, July 2021.
[7] Jyoti, Kirti Bhatia, Rohini Sharma,
Development of Wavelet Based Image

Denoising, International Journal of Trend in
Scientific Research and Development, Volume
4 Issue 5, July-August 2020.
[8] Jyoti, Kirti Bhatia, Shalini Bhadola, Rohini
Sharma, An Analysis of Facsimile
Demosaicing Procedures, International journal
of Innovative Research in computer and
communication engineering, Vol-08, Issue-
07,july 2020.
[9] Jonathan, M. (2005) Black Ledge. Digital
Image Processing Mathematical and
Computational Methods Coll House,
Watergate, Chic Ester, West Sussex, PO20
3QL, England.
[10] Newman, Richard (2011). Conservation and
care of museum collections (1st ed.). MFA
publications. p. 29. ISBN 978-0-87846-729-7.
[11] Idelson, Antonio; Severini, Leonardo (28 June
2018). "Inpainting". The Encyclopedia of
Archeological Sciences: 1–4, October 30, 2019.
[12] Michal Aharon; Michael Elad; Alfred
Bruckstein (2006), "K-SVD: An Algorithm for
Designing Over complete Dictionaries for
Sparse Representation" (PDF), IEEE
Transactions on Signal Processing, 54 (11):
4311–4322.
[13] Rubinstein, R., Bruckstein, A.M., and Elad, M.
(2010), "Dictionaries for Sparse Representation
Modeling", Proceedings of the IEEE, 98 (6):
1045–1057.
[14] Etienam, Clement, 4D Seismic History
Matching Incorporating Unsupervised
Learning, 2019.
[15] G. David, S. Mallat, and Z. Zhang, Adaptive
greedy approximations," J. CONSTRUCT.
APPROX, VOL.13, 1997

Development and Analysis of Enhanced Image Inpainting Approach

More Related Content

Similar to Development and Analysis of Enhanced Image Inpainting Approach (20)

More from ijtsrd (20)

Recently uploaded (20)

Development and Analysis of Enhanced Image Inpainting Approach