IRJET- Image De-Blurring using Blind De-Convolution Algorithm

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 07 | July 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 894
Image De-blurring using Blind De-convolution Algorithm
P. P. Ghugare1, S. S. Thorat2
1P.G. Scholar, Dept. of Electronics Engineering, GCOE, Amravati
2Assitant Professor, Dept. of Electronics Engineering, GCOE, Amravati
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - Blurring is usually caused by defocus or
relative motion, which can be formulated by the convolution
of the point-spread-function (PSF) and latent image.
Advanced techniques such as image enhancement, de-
blurring, denoise, and super resolution have been developed
to improve image quality post-digitization. De-convolution
using blind method is very complex method because
recovery of image is performed using less or no prior
knowledge of the point spread function (PSF). In this
propose method blind de-convolution technique has been
implemented to de-blur a image. To add blur in the image
the different blur models are used here, these are Gaussian
Blur, Motion Blur, Average blur. Maximum likelihood
estimation technique is utilized using our propose blind de-
convolution algorithm for better results. The various
parameters have been calculated such as RMSE, MSE, PSNR,
SNR, SSIM and Histogram. Experimental results show that
blind de-convolution method runs faster than conventional
work.
Key Words: Blur, Image Restoration, Image degradation,
Deblurring, PSF
1. INTRODUCTION
In daily life, many images such as photographs,
pictures, books, video and so on, so the image and human
life are indivisible. With the fast growth in modern digital
technology, using digital image as digital information
carrier has been the people's attention. The digital images
are used in various area, such as medical, military and
transportation, microscopy imaging and photography
deblurring etc. The recorded image consisting a noise and
blur version of original picture. The analysis of various
pictures using techniques that can identify shades, colours
and relationships which cannot be perceived by the
human eye.
Image deblurring is an inverse problem whose aspire
is to recover an image which has suffered from linear
degradation. The blurring degradation can be space
variant or space-in variant. Image deblurring methods can
be divided into two classes: Non-blind, in which the
blurring operator is known and blind, in which the
blurring operator is unknown. Blurring is a form of
bandwidth reduction of the image due to imperfect image
formation process. It can be caused by relative motion
between camera and original image. Normally, an image
can be degraded using low-pass filters and its noise. This
low-pass filter is used to blur/smooth the image using
certain functions. Image restoration is to improve the
quality of the degraded image. It is the process of
recovering the original scene image from a degraded or
observed image using knowledge about its nature. Blind
Image De-convolution is a more difficult image restoration
where image recovery is performed with little or no prior
knowledge of the degrading PSF.
In this paper, the new blind de-convolution
algorithm is designed to restore a original image from the
degraded image. The main objective of this paper is to
restore a original image from degraded image. This paper
is structured as follows: Section 2 describes the literature
survey on restoration of an image. Section 3 describes the
deblurring algorithm and overall architecture of this
paper. Section 4 describes the sample results for de-
blurred images using our proposed algorithm. Section 5
describes the conclusion.
2. LITERATURE REVIEW
Previous related researches have worked on the
various filtering techniques to reduce noise and blur factor
of image, but it has its own disadvantages and then
developed various deblurring algorithms. Various work
has been done on deblurring algorithms on various
platform and with various assumption. Some related
proposed work has been discuss below.
Satoshi Motohashi [1] worked on gradient reliability
map(R-map). In this paper a novel algorithm based on
two-step blind de-convolution is done. In this paper,
during latent image restoration step, total variation
regularization is applied to reduce texture components
and noise; and shock filter is applied to emphasize the
edges. The gradient reliability map is then applied to
decrease the edges, which are severely affected in the PSF
estimation. Fu-Wen Yang [2] proposed a algorithm on the
blind deblurring method. In this blind deblurring method
needing to predict a blur kernel in own way. The color
distribution of edges is more distinct in clear image than in
a blurred image. The filter is proposed to make edges in a
blurred image clearer for use as reference image.
Marapareddy. R [3] worked on Wiener filtering for blur
image restoration which is degreded due to complex
surrounding environment. Here first find out atmospheric
turbulence degradation model. After that inverse filtering
and minimum mean square error i.e., for restore the
blurring image the wiener filtering is applied. Shuyin Tao
[4] formulate the de-convolution problem by combining

negative logarithmic poisson likelihood with total
variation (TV) regularization, and describe a fast
algorithm which is based on the method of Lagrange
multiplier to solve it. The restored image is achieved by
alternately solving two sub-problems. Rinku kalotra [6]
worked on the two popular restoration techniques viz.
LRA and BID are used and analyzed in the restoration of X-
ray images. X-ray image play a important role in
considering with the detection of several disease in a
patients and they face the problem of motion blur and
Gaussian noise. Satoshi Hirano [5] worked on the blind
method restoration that rapidly restored blurred image
using local patches. In this algorithm, a portion of blurred
image is utilized for the PSF(Point Spread Function)
calculation. In addition, a new technique proposed for a
automatic PSF size estimation algorithm which is used for
an generation of autocorrelation map. Punam patil [8]
worked on the blind de-convolution technique using canny
edge detector. In the blurred image edges the canny edge
method is used for detection of ringing effect and then it
can be removed before restoration process. Masanao
Sawada [7] worked on novel blind image restoration
algorithm which is depends on the total variation(TV)
regularization and the shock filter. It consist of alternative
iteration of the point spread function calculation and de-
convolution.
3. METHODOLOGY
The image restoration techniques is used to reduce noise
and recover resolution loss. Image processing techniques
are performed either in the image domain or in the
frequency domain. The general system design describes
the blind image restoration is represented in a block
diagram given below in Figure 1
Fig -1: Block Diagram for Blind Image Restoration
3.1 Degradation Model
There are varieties of degradation models are studied that
contribute to the recording of a distorted image. Although,
there are non-linear degradation models, where the
distortion becomes a function of the GTI itself, such as X-
ray images. In this work, linear models, which are
reasonable to model many distortions in photography. The
simplest model takes the form of a result of two
phenomena, namely degradation due to image acquisition
or defects of the imaging system and distortion due to
random noise. Mathematically, the model accounting for
this degradation can be represented as,
b = k g + n
Where b is the distorted image, k is a linear operator and n
represents the additive random noise. In image
restoration, k is mostly referred as the point spread
function (PSF) or blur kernel. An ideal imaging system
would have a delta function as the PSF, which however is
not realistic.
Fig -2: Degradation Model
3.2 Restoration Model
In Restoration model, the degraded image is reconstructed
using restoration filters. In this process noise and blur
factor is removed and get an estimate of the original image
as a result of restoration. The closer the estimated image is
to the original image the more efficient is our restoration
filter. Figure 3 below represents the structure of
restoration model.
Fig -3: Restoration Model

There are numerous techniques and algorithms available
for Image restoration. Each technique has its own features.
In this method the new blind de-convolution technique is
used. This Technique allows the reconstruction of original
images from degraded images even when we have very
little or no knowledge about PSF. These techniques are
more difficult to implement and are more complicated as
compared to other category.
3.3 Overall Architecture of Image Restoration
Fig -4: Overall Architecture of Image Restoration
This section presents the proposed work for Blind
Image Restoration technique. In particular, this section is
dividing into two sections, which deal with the addition
blur and noise in image with different blur models and the
next restoration of the image with blind de-convolution
technique. Figure 5 shows flow chart of restoration
process which describe the whole process step by step of
de-blurring algorithm. In this propose algorithm the main
aim is to reconstruct original image from the degraded
image. First take the original image. Convert this image
into gray scale image. After Converting into gray scale
resize the image.
After addition of degradation function such as Gaussian
blur, Average blur, Motion blur in original image blurred
image is formed. After that Gaussian noise is added into
the degraded image. Analyze the de-blur using undersize
PSF, oversize PSF and initial PSF. Then blind de-
convolution algorithm is applied for better reconstruction
of the PSF and for obtaining the original image. Then
calculate different parameters.
4. EXPERIMENTAL RESULTS
This provides an overall performance of the experimental
models based on blind de-convolution algorithm. In this
blind de-convolution is performed on three different
models. The experiment is performed on standard dataset
of MATLAB.
4.1 Gaussian Blur Model
In the Gaussian blur model for performing the operation
first take the original image of size 526 by 526 as an input
image. The input image is resized into 256 by 256 size. The
color image consists of three planes, so the preprocessing
on color image is difficult. Hence original image is
converted into gray scale. In the gray scale image the
Gaussian blur is added in the image. The Gaussian blur
creates degradation in the image. After addition of blur
some parts of information of image is lost in processing of
image. After preprocessing the image obtain is blur image.
After addition of blur the Gaussian noise is added in the
image. This Gaussian noise creates unwanted variation of
brightness. The Gaussian noise creates disturbance in the
image. On the blur and noise image the blind de-
convolution algorithm is performed with PSF. Thus de-
blur the image to obtain the original image.
Fig -5: Blur Image

Fig -6: Deblur with PSF
4.2 Motion Blur Model
In the original gray scale image the motion blur is added in
the image. The motion blur creates degradation in the
image. After addition of blur some parts of information of
image is lost in processing of image. The Gaussian noise is
added in the image. This Gaussian noise creates unwanted
variation of brightness and also creates disturbance in the
image. On the blur and noise image the blind de-
convolution algorithm is performed with PSF. Thus de-
blur the image to obtain the original image.
Fig -7: Motion Blur Image
4.3 Average Blur Model
In the original gray scale image the average blur is added
in the image. The Gaussian blur creates degradation in the
image. After addition of blur some parts of information of
image is lost in processing of image. On the blur and noise
image the blind de-convolution algorithm is performed
with PSF. Thus de-blur the image to obtain the original
image.
Fig -9: Average Blur Image
We used images to experimentally validate the
proposed method. Table I shows the experimental
parameters. The above figure shows the experimental
results of a blurred image and the results of the proposed
method, respectively. Table shows the computation time
of the proposed method for the experimental images. The
results show that it is possible to obtain sharp images
because the correct PSF can be estimated. Moreover, the
reconstructed images of the proposed method are clearer
than those of the conventional method.

Table -1: Gaussian Blur Model
Parameters Conventional Method
(Lucy-Richardson)
Proposed Method
(Blind De-convolution)
MSE 32.6753 33.4153
RMSE 2.40759 2.40429
PSNR 31.5413 32.8913
Correlation 0.96853 0.97854
SNR 8.4429 8.4429
SSIM 0.7283 0.7269
Table -2: Motion Blur Model
(Lucy-Richardson)
Proposed Method
MSE 43.8702 42.4202
RMSE 2.5920 2.5520
PSNR 29.4321 31.8551
SNR 7.0204 7.5204
SSIM 0.6448 0.6478
Table -3: Average Blur Model
(Lucy-Richardson)
Proposed Method
MSE 42.8752 40.3342
RMSE 2.5178 2.5207
PSNR 28.5931 29.6532
SNR 7.3665 7.3617
SSIM 0.6945 0.6961
The results gives in table above for the three
different blur models. The result shows that the different
models give the different parameters. In the above table
the Gaussian blur model shows the best PSNR value with
less MSE value. In this system the blind de-convolution
algorithm is performed on different image for finding the
parameters. This output parameters is compared with the
conventional method i.e Lucy-Richardson.
Table -4: Processing Time
Image Size Conventional Method
(Second)
Proposed Method
(Second)
482 × 482 12.31 9.02
1920 × 1080 13.05 10.11
753 × 502 12.16 9.19
1024 × 683 13.21 11.01
The timing comparison table shows that how much
timing is required to run the images. This method runs
faster than the conventional method.
5. CONCLUSION
This present study demonstrated the various
restoration techniques that have been developed to
restore the original image from the degraded image. The
proposed schemes along with a number of restored
schemes are simulated on standard and naturally blurred
images under different parametric PSFs. Here mainly
three techniques have been simulated and compared. The
results show the performance by the gaussian blur, motion
blur and average blur. This algorithm have different
results and the performance of these algorithm is
measured in terms of parameter like PSNR (Peak Signal to
Noise Ratio), MSE (Mean Square Error), SNR (Signal to
Noise Ratio), RMSE (Root Mean Square Error), SSIM and
Histogram. Higher the value of PSNR shows the more clear
image quality. Experimental results demonstrate that our
method runs an order of magnitude faster than previous
work, while the deblurring quality is comparable. GPU
implementation makes method fast enough for
experimental use.
REFERENCES
[1] S. Motohashi, T. Nagata, T. Goto, Reo Aoki, Haifeng
Chen, “A Study on Blind Image Restoration of
Blurred Images using R-map", IEEE International
Workshop on advanced image technology
(IWAIT), 2018.
[2] Fu-Wen Yang, Hwei Jen Lin, and Hua Chuang,
“Image Deblurring", IEEE Smart World,
Ubiquitous Intelligence and Computing, Advanced
and Trusted Computed, Scalable Computing and
Communications, Cloud and Big Data Computing,
Internet of People and Smart City Innovation,
2017.
[3] Marapareddy. R, “Restoration of blurred images
using wiener filtering", International Journal Of
Electrical, Electronics and Data Communication,
2017.
[4] S. Tao, W. dong, Q. Wang, Z. Tang, “Fast Non-blind
De-convolution Method for Blurred Image
Corrupted by Poisson Noise", IEEE International
conference on Progress in Information and
Computing (PIC), 2017.
[5] S. Hirano, M. Sakurai, T. Goto, H. Sensshiki, “Fast
Blind Restoration of Blurred Images Based on
Local Patches", IEEE International Conference on
Electronics, Information, and Communications
(ICEIC), 2016.

[6] Li Yang, “Image Restoration from a Single Blurred
Photograph", IEEE 3rd International Conference
on Information Science and Control Engineering,
2016.
[7] Van Ge, “Research on the blind restoration
algorithm of motion-blurred image", IEEE
Advanced Information Management,
Communicates, Electronic and Automation
Control Conference (IMCEC), 2016.
[8] Feng Zeng, Wei Wang, “Exposing Blurred Image
Forgeries through Blind Image Restoration", IEEE
10th International Conference on P2P, Parallel,
Grid, Cloud and Internet Computing (3PGCIC),
2015.
[9] Rinku Kalotra, Sh. Anil Sagar, “A Novel Algorithm
for Blurred Image Restoration in the field of
Medical Imaging", International Journal of
Advanced Research in Computer and
Communication Engineering, 2014.
[10] K. Ohkoshi, M. Sawada, T. Goto, S. Hirano, and M.
Sakurai, “Blind Image Restoration Based on Total
Variation Regularization and Shock Filter for
Blurred Images". IEEE International conferene on
Consumer Electronics, pp. 219-220, 2014.
[11] H. Riyaz Fathima, Mr. K. Madhan Kumar,
“Evaluation of Blind image Restoration",
International Journal of Advanced Research in
Computer Engineering and Technology (IJARCET),
pages: 4210-4215, 2014.
[12] P. Patil, R .B. Wagh, “Implementation of
Restoration Of Blurred Image Using Blind
Deconvolution Algorithm", International
Conference on Wireless and Optical
Communications Networks (WOCN), 2013.
[13] P. Jaypriya, Dr. R. M. Chezhian. “A Study on Image
Restoration and its Various Blind Image
Deconvolution Algorithms". International Journal
of Computer Science and Mobile Computing, Page:
273-278, 2013.
[14] Aftab Khan, Hujun Yin, “Quality measures for
blind image deblurring", IEEE International
Conference on Imaging Systems and Techniques
Proceedings, 2012.
[15] Fan Fan, Kecheng Yang, “Application of Blind De-
convolution Approach with Image Quality Metric
in Underwater Image Restoration", IEEE
International Conference on Image Analysis and
Signal Processing, 2010.
[16] Gopal Harikumar, Y. Bresler, “Perfect Blind
Restoration of Images Blurred by Multiple Filters:
Theory and Efficient Algorithms", IEEE
Transactions on Image Processing, Feb 1999.

IRJET- Image De-Blurring using Blind De-Convolution Algorithm

More Related Content

What's hot (19)

Similar to IRJET- Image De-Blurring using Blind De-Convolution Algorithm (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Image De-Blurring using Blind De-Convolution Algorithm