I J STEI nternational J ournal of Science Technology & Engineering| Vol.

1, I ssue 1, J uly 2014| I SSN(online): 2349-784X

All rights reserved by www.ijste.org
11
An Image Inpainting Using Patch-Based
Synthesis Via Sparse Representation

Nirali Pandya Mayank Pandya
Student Hardware and Networking manager
Department of Computer Science and Engineering Department of Hardware and Networking
Parul group of institute, Gujarat technical university Magnum Company PVT. LTD

Abstract

Image inpainting is a art of missing value or a data in an image. The purpose of image inpainting is to reconstruct missing regions
which is visible for human eyes. Image inpainting is the process of reconstructing lost part of images based on the background
information. Image inpainting is a technique for restoring damaged old photographs and removing undesired objects from an image.
The basic idea behind the technique is to automatically fill in lost or missing parts of an image using information from the
surrounding area. It is used for restoration of old films and object removal in digital photographs. It is also applied to red-eye
correction, stamped data from photographs, dust spot in film, removing objects to creative effect etc. The main goal of the Inpainting
algorithm is to modify the damaged region in an image. In this paper we provide a review of different techniques used for image
Inpainting. We discuss texture synthesis method and inpaint the image using masking.
Keywords: Image inpainting, Texture Synthesis, Exemplar, Object Removal, PDE based image inpainting, sparse
representation.
_________________________________________________________________________________________________________
I. INTRODUCTION
Inpainting is the art of restoring lost parts of an image and reconstructing them based on the background information. This has to be
done in an undetectable way. The term inpainting is derived from the ancient art of restoring image by professional image restorers in
museums etc. Digital Image Inpainting tries to imitate this process and perform the inpainting automatically. The filling of lost
information is essential in image processing, with applications as well as image coding and wireless image transmission, special
effects and image restoration. The basic idea at the back of the algorithms that have been proposed in the literature is to fill-in these
regions with available information from their environment [5].
In this paper, different types of image inpainting techniques presented are discussed in Section 2, discuss about the patch-based
texture synthesis method in Section 3 discuss about experimental results in section 4 and the conclusion in last section.
II. SURVEY ON IMAGE INPAINTING TECHNIQUES
There are mainly two methods found in literature of inpainting:
Structural inpainting
Textural inpainting
Structural inpainting uses geometric approaches for filling in the missing information in region. This algorithm focus on the
consistency of the geometric structure. This algorithm also used in the decomposition is of bounded variation, representing the
underlying image structure [4].
Textural inpainting use to captures the texture and possible noise and also complete the missing regions using similar
neighborhoods of the damaged pixels [4].
Diffusion based Inpainting was the first digital Inpainting approach. In this approach missing region is filled by diffusing the
image information from the known region into the missing region at the pixel level. The drawback of the diffusion process is it
introduce s some blur, which becomes noticeable when filling larger regions [3].
PDE based algorithm, this algorithm is the iterative algorithm. PDE based technique has been widely used in number of
applications such as image segmentation. If missing regions are small one then this algorithm is works a good result but when the
missed regions are large this algorithm will take so long time and it will not produce good results. Another category of Inpainting is
exemplar- based Inpainting algorithm. This method of image Inpainting is an efficient approach to reconstructing large target
regions. Exemplar-based Inpainting approach iteratively synthesizes the target region by most similar patch in the source region.
These algorithms also overcome the drawbacks of PDE based inpainting. Also it removes smooth effect of the diffusion based
Inpainting algorithm [3].
Generally, an exemplar-based Inpainting algorithm includes the following four main steps:
An Image Inpainting Using Patch-Based Synthesis Via Sparse Representation
(IJSTE/ Volume 01/ Issue 01 / 002)


All rights reserved by www.ijste.org
12

Fig. 1: Structure propagation by exemplar-based texture synthesis.(a) Original image, with the target region , its contour and the source region
clearly marked.

(b) We want to synthesize the area delimited by the patch p centred on the point p .[2]
Initializing the Target Region, in which the initial missing areas are extracted and represented with appropriate data
structures.
Computing Filling Priorities, in this a predefined priority function is used to compute the filling order for all unfilled pixels
p in the beginning of each filling iteration.
Searching Example and Compositing, in which the most similar example is searched from the source region to compose
the given patch, (of size N N pixels) that centered on the given pixel p.
Updating Image Information, in which the boundary of the target region and the required information for computing
filling priorities are updated.
III. TEXTURE SYNTHESIS METHOD
The Texture Synthesis method are used to complete the missing regions using similar neighbourhoods of the damaged pixels. The
texture synthesis algorithms synthesize the new image pixels from an initial seed. but depends on user choices on sampling location
and content.
Texture synthesis approaches can be categorized into
Three categories:
Statistical (parametric)
pixel-based (non-parametric)
patch-based (non-parametric)
Statistical methods are more likely to succeed in reproducing stochastic / irregular textures, but usually fail to reproduce
structured/regular textures. pixel-based methods build on the sample texture pixel-by-pixel instead of applying filters on it, and
their final outputs are of better quality than those of statistical methods, but they usually fail to grow large structured textures.patch-
based [7]methods build on a sample texture patch-by-patch as opposed to pixel-by-pixel, thus they yield faster and more plausible
regular textures.
Recently, a study for patch-based texture synthesis algorithms has shown that for handling special types of texture we have to
develop the special purpose algorithms. Taking this aim and the variety of algorithms for texture synthesis into consideration, we
can conclude that there is no universal texture synthesizer is present. Still it remains a goal to be desire.Why we have to use patch-
based texture synthesis algorithm instead of pixel-based algorithm explain as follows: [8]

Fig. 2: Comparison between pixel and patch.
In pixel-based algorithm, the copy is just a copy.
However, in patch-based algorithms, the issue is more complicated as a patch, being larger than a pixel, usually overlaps with the
already synthesized portions, so some decision has to be made about how to handle the conflicting regions. new patches simply
overwrite over existing regions. By using patches with irregular shapes, this approach took advantage of the texture masking effects
of human visual system. [3]
An Image Inpainting Using Patch-Based Synthesis Via Sparse Representation
(IJSTE/ Volume 01/ Issue 01 / 002)


All rights reserved by www.ijste.org
13
Problem behind this concept is overlapping the selected patches in the image.
Using Texture Synthesis method we can select the patches and also handle the conflicting region with the overlapping the patches.
Input the image
Selecting the target region
Find the boundary of the target region
Selecting the patch
finding patch priority
Matching with each similar patches including different pixels
use texture synthesis method for manage the patches
Do the Linear combination with each patches using matrix
For using Sparsity create one redundant dictionary
After generating the dictionary need to create the signal for Sparsity
X=D*
Where X=signal D=redundant dictionary = non-zero vector
Fill the hole using sparse representation
Get the inpainted image
Work flow is mention above first select the patch with the boundary that is target region which we have to inpaint. Overlapping
with two patches is handling by texture synthesis method and then calculate the max priority patch with the best exemplar algorithm.
IV. EXPERIMENTAL RESULTS
Here, Implementation result is explained in brief. In Figure4 (a) First image is corrupted image that is full of additive noise and blur
from whole image we have to remove that blur with the inpainting methods, for that we have to generate the dictionary including
number of patches then match all the patches according to their priority with the similar patches and inpaint that portion with perfect
similar patch and we get the output that is given in last image. Second Image is Original Image. In Figure4(b),is The dictionary
trained on the corrupted image. Dictionary generate with the number of patches.

Fig. 4(a): Implementation work

Fig. 4(b): The dictionary trained on the corrupted image

Figure 5(A): Implementation Results
An Image Inpainting Using Patch-Based Synthesis Via Sparse Representation
(IJSTE/ Volume 01/ Issue 01 / 002)


All rights reserved by www.ijste.org
14

Fig. 5(B): The dictionary trained on the corrupted image

Fig. 6(a): Implementation work

Fig. 6(B): The dictionary trained on the corrupted image
V. CONCLUSION AND FUTURE WORK
A famous texture synthesis based inpainting algorithm proposed by [2] selects the most suitable patch from the source region at each
step, thus it can be viewed as a special case of our algorithm if we constrain that only one patch is selected in each step. texture
synthesis is actually a sparse representation method with the constraint that L0 norm (the number of nonzero entity in vector) equals
to 1. Thus, the gap between texture synthesis and sparse representation is bridged. However, in general, our algorithm selects several
columns of the dictionary, i.e. several different patches if we construct the dictionary by directly sampling patches, and use a linear
combination of them to fit the target patch. Thus, our algorithm has higher capability to fit diverse patches for its form of linear
combination.
REFERENCES
[1] Anupam , Pulkit Goyal and Sapan Diwakar, Fast and Enhanced Algorithm for Exemplar Based Image Inpainting,Information Technology .
[2] A. Criminisi, P. Perez and K. Toyama, Object Removal by Exemplar-Based Inpainting, Microsoft Research Ltd.,2003,pp.1-8
[3] Komal s Mahajan, Prof. M. B. Vaidya, Image in Painting Techniques: A survey, IOSR Journal of Computer Engineering, 2012 ,Vol.5(4),pp.45-49
[4] Marcelo Bertalmio, Luminita Vese, Guillermo Sapiro, Simultaneous Structure and Texture Image Inpainting, august 2003, ieee transactions on image
processing, vol. 12, no. 8,
[5] V. Janarthanan,G. Jananii, A Detailed Survey on Various Image Inpainting Techniques, September 2012 , International Journal of Advances in Image
Processing, Vol. 2, No. 3
[6] J. C. Huang, C. H. Hwang, Y. C. Liao, N. C. Tang, and T. J. Chen, Exemplar-based image inpainting base on structure.
[7] [7] A. Efors and T. Leung, Texture synthesis by
[8] non-parametric sampling, in Proceedings of the 17th IEEE International Conference on Computer Vision, 1999, pp. 1033-1038.
[9] [8] Li-Yi Wei, Sylvain Lefebvre,Vivek Kwatra,
[10] Greg Turk, A State of the Art in Example-based Texture Synthesis,EUROGRAPHICS 2009.