Image Restoration Based On Deconvolution by Richardson Lucy Algorithm

International Journal of Engineering Trends and Technology (IJETT) Volume 14 Number 4 Aug 2014
Image Restoration Based On Deconvolution by

Richardson Lucy Algorithm
Madri Thakur#1, Shilpa Datar*2
#1
PG student, Dept. of Electronics and Communication Engg.

SATI, Vidisha, M.P., India
*2
Asst. Prof., Dept. of Electronics and Instrumentation Engg
SATI, Vidisha, M.P., India
Abstract This article presents the performance analysis of

different basic techniques used for the image restoration.
Restoration is a process by which an image suffering from
degradation can be recovered to its original form. Removing blur
and noise from image is very difficult problem to solve. We have
implemented the three different techniques of image restoration
and tested our implementation for the blurred image in the
standard environment. We have obtained the blurred image with
the standard blurring functions and the noise. The degraded
images have been restored by the use of Wiener deconvolution,
Inverse deconvolution and
RichardsonLucy algorithm.
Further we have compared the different results on the basis of
PSNR and MSE values of the restored image. Finally the
conclusion is formulated.
image restoration. Often the blur kernel is assumed

to be space-invariant [4-5]. If we lack prior
knowledge of the blur kernel or point spread
function h, we have the more difficult blind (linear)
image restoration problem in which h also needs to
be estimated. Also we have
G (u) = F(u)H(u) +N(u)

The
(2)
are the Fourier transform of g, f, h, .
Keywords Inverse filter, Wiener filter, Lucy- Richardson, MSE

(mean square error), PSNR (peak signal to noise ratio).
1. INTRODUCTION
The task of deblurring, a form of image restoration,
is to obtain the original, sharp version of a blurred
image.[1-3] There exist many applications for
image restoration, including astronomical imaging,
medical imaging, law enforcement, and digital
media restoration. The problem has attracted strong
research interest and will continue to do so, not
only because it has many applications but also
because it has a simple mathematical formulation
yet it is a classical inverse problem for which good
solutions are not easily obtained. The simple
equation for expressing image blurring/degradation
is as follows;
g=f*h+
(1)
Where f is the original image and g is the version
that has been degraded through blurring
(convolution ) by kernel h and the addition of
random noise . This degradation model represents
a linear relationship between f and g; hence, the
problem of recovering f from g is called linear
ISSN: 2231-5381
Fig. 1 Image Blurring model
Image deconvolution methods are used to

estimate the latent image from the degraded image.
They can be divided into two categories, non-blind
and blind deconvolution [6-7]. In non-blind
deconvolution, the PSF is known and f can be
restored through an error.
Minimization process. The Weiner deconvolution
and the Inverse deconvolution are the two
commonly used methods, within this category.
RichardsonLucy
deconvolution
is
algorithm which is based on iteration process. Its
performance in the presence of noise is found
to be superior to that of other deconvolution
algorithms[8-9].
http://www.ijettjournal.org
Page 161
3. INVERSE DECONVOLUTION METHOD

FOR IMAGE DEBLURRING
Direct inverse Filtering is the simplest approach to
restoration [9]. In this method, an estimate of the
Fourier transform of the image (u, v) is computed
by dividing the Fourier transform of the degraded
image by the Fourier transform of the degradation
function
(u, v) = G (u, v) / H (u, v)
(3)
This method works well when there is no additive
noise in the degraded image. That is, when the
degraded image is given by
g(x, y) = f(x, y)*h(x, y)
(4)
But if noise gets added to the degraded image

then the result of direct inverse Filtering is very
poor. Equation 1.gives the expression for g(u, v).
Substituting for G(u, v) in the above equation, we
get
(u, v) = F ((u, v) + N(u, v)H(u, v)
(5)
The above equation shows that direct

inverse Filtering fails when additive noise is present
in the degraded image. Because noise is random
and so we cannot find the noise spectrum
.
Fig. 2 Inverse Deconvolution
ISSN: 2231-5381
4. WEINER DECONVOLUTION METHOD

FOR IMAGE DEBLURRING
Weiner deconvolution is named after Norbert
Weiner, who first proposed the method in 1942 w
Weiner filtering is one of the earliest and best
known approaches to linear image restoration [9].
Weiner Filtering is more robust in the
presence of additive noise. Weiner filtering
incorporates both degradation function and
statistical characteristics of noise into the
restoration process. The objective of this technique
is to find an estimate of the original image f such
that the mean square error between them is
minimized. This error measure is given by
e 2 E{(f - f ) 2 }
(6)
Where E{.} is the expected value of the argument.

The method is founded on considering image and
noise as random processes and objective is to find
an estimate of the uncorrupted image such that
the mean square error between them is minimized.
If the noise is zero, then the noise power spectrum
vanishes and the wiener filter reduces to the inverse
filter.
5. RICHARDSONLUCY DECONVOLUTION
ALGORITHM
The non blind de-convolution is the category of deconvolution method in which the PSF is known.
The RichardsonLucy deconvolution algorithm has
become popular in the fields of astronomy and
medical imaging. Initially it was derived from
Bayes theorem in the early 1970s by Richardson
and Lucy. In the early 1980s it was redeliver by
Shepp and Vardi as an algorithm to solve positron
emission tomography imaging problems, in which
Poisoning statistics are dominant. Their method
used a maximum-likelihood solution, which was
found by use of the expectation maximization
algorithm of Dempster et al
Page 162
The reason for the popularity of the

RichardsonLucy algorithm is its implementation
of maximum likelihood and its apparent ability to
produce reconstructed images of good quality in the
presence of high noise levels. We therefore
assumed that a non blind form of this algorithm
would have the same characteristics [2-6].
Where is the convolution operation. F= is the

estimate of the un degraded image. We have used
the R-L algorithm iteratively staring from the
blurred image. The image restored after each
iteration moves closer to the original image thus
reducing MSE with number of iterations. The
program execution is terminated when MSE
obtained becomes constant in consecutive iterations
Non linear iterative technique is better than the

linear technique. Non linear behaviors is not always
7. PERFORMANCE PARAMETERS
predictable and required computational resourcesR algorithm which is arise from the maximum like Image restoration research aims to restored image
hood formulation
to from a blurred and noisy
image
^
^
^
A
widely
used
measure
of
reconstructed
image
f k 1 (x, y) f k (x, y)[h(-x,-y) * g(x, y)/h(x, y)* f k (x, y)]
fidelity for an N * M size image is the mean square
(7)
error (MSE) and is given by
6. RESULTS AND ANALYSIS

We have evaluated the results for different images .The results are shown for the two images for the
different variance and image sizes.
Table 1 Results for the Cameraman image
Image size
256x256
Noise variance
512x512
0.05
0.007
0.05
0.007
Inverse filter
MSE
0.0262
PSNR
15.8123
MSE
0.0068
PSNR
21.7024
MSE
0.0042
PSNR
22.7645
MSE
0.0083
PSNR
20.7954
Wiener filter
0.0138
18.5958
0.0061
22.1819
0.0020
27.0082
0.0028
25.5632
Iteration 1
0.0680
11.6762
0.0654
11.8434
0.0210
16.7799
0.0440
13.5639
Iteration 10
0.0268
15.7245
0.0245
16.1020
0.0019
27.1238
0.0064
21.9615
Iteration 20
0.0112
19.5254
0.0071
21.5087
0.0011
29.6501
0.0022
26.6454
Iteration 30
0.0070
21.5233
0.0035
24.6023
8.9311e004
30.4910
0.0018
27.4981
Richardson lucy
Table 2 Results for the peppers image

Image size
256x256
Noise variance
512x512
0.007
0.05
0.007
Winer filter
Iteration 1
MSE
0.0200
0.0052
0.0402
PSNR
16.9817
22.8473
13.9617
MSE
0.0045
0.0061
0.0307
PSNR
23.4626
22.1819
15.1323
MSE
0.0202
0.0057
0.0270
PSNR
16.9483
22.4348
15.6885
MSE
0.0042
0.0020
0.0210
PSNR
22.7645
27.0082
16.7799
Iteration 10
0.0046
23.3504
0.0021
26.8659
0.0033
24.8779
0.0019
27.1238
Iteration 20
0.0032
24.9371
0.0012
29.3916
0.0024
26.1250
0.0011
29.6501
Iteration 30
0.0028
25.5736
9.0215e-004,
30.4472
0.0025
25.9579
8.9311e-004
30.4910
Inverse filter
Richardson lucy
0.05
ISSN: 2231-5381
Page 163
original image
original image
image affected by blurring
image affected by blurring
50
50
100
100
150
150
200
200
250
(a)
250
(a)
50
image restore by weiner filtering
50
50
100
100
150
150
200
200
250
250
50
image restore by Inverse filtering
(b)
100
150
image restore by Inverse filtering
200
50
50
100
100
150
150
200
200
50
(b)
100
150
200
250
50
restored image after the itrretion 30
(d)
100
150
200
100
150
image restore by weiner filtering
200
250
200
250
250
250
50
(b)
250
100
(b)
150
(d)
50
200
250
restored image after the itrretion 30
100
150
250
50
50
100
100
150
150
200
200
250
50
250
50
100
150
(e)
200
250
Fig.3 Results of pepper.png (a) original image (b) blurred image (c)
Restored by Inverse filter (d) Restored by Wiener filter (e)Restored
by R-L at iteration 30.
From Fig.(3) (c) & (d), and Fig.4 (c) & (d), the
above results we found that the inverse filter works
better than the Weiner filter, under noise conditions.
When the variance of noise increases the
performance of inverse filtering not provides the
sufficient PSNR. The Weiner filtering gives the
good PSNR regardless of the noise variance
Form Fig.3 (e) and 4 (e), the results obtained by the
Richardson Lucy method, we have found that the
PSNR increases with the number of iterations and
also the quality of the image enhances.(See Table 1
and 2).
ISSN: 2231-5381
100
(e)
150
200
250
Fig.4 Results of cameraman.tif (a) original image (b) blurred image

(c) Restored by Inverse filter (d) Restored by Wiener filter
(e)Restored by R-L at iteration 30.
8. CONCLUSION
We have seen the requirement and significance of
image de-blurring. We have seen the mathematical
formulation for the blurred image. we already have
the knowledge of point spread function .
Weiner filtering provides the better results than the
inverse filtering almost in every condition except
when the noise having very less variance.
The Richardson Lucy provides good estimate for
the blurring function and gives better PSNR within
the limited iterations. Yet if we use this method
with the known point spreading function then it is a
time taking method, still it can provides the PSNR
even better than Weiner deconvolution.
With the help of the basic
method of deconvolution, we may try to form some
deconvolution method which can provide better
Page 164
PSNR within the very less iterations for the blind

deconvolution.
REFERENCES
[1]
J. G. Walker, D. A. Fish, A. M. Brinicombe, and E. R. Pike, Blind

deconvolution by means of the RichardsonLucy algorithm J. Opt. Soc.
Am. A, Vol. 12, No. 1, January 1995.
[2]
Arijit Dutta, Aurindam Dhar, Kaustav Nandy, Project report on Image

Deconvolution By Richardson Lucy Algorithm, Indian Statistical
Institute, November, 2010.
[3]
G. R. Ayers and J. C. Dainty, Iterative blind deconvolution method and

its applications, Opt. Lett. Vol.13, 547549 , 1988 .
[4]
J. R. einup, Phase retrieval algorithms: a comparison, Appl. Opt.

Vol.21, 27582769, 1982.
[5]
B. L. K. Davey, R. G. Lane, and R. H. T. Bates, Blind deconvolution of

noisy complex-valued image, Opt. Commun. Vol.69, 353356, 1989.
[6]
W. H. Richardson, Bayesian-based iterative method of image

restoration, J. Opt. Soc. Am. Vol.62, 5559 1972.
[7]
L. B. Lucy, An iterative technique for the rectification of observed

distributions, Astron. J. Vol. 79, 745754, 1974.
[8] Ramesh Neelamani, Thesis report on Inverse problems in image

processing Electrical and Computer Engineering Rice University,
Houston, Texas.
[9] Reginald L. Lagendijk and Jan Biemond, Basic methods for image
restoration and Identification, Lagendijk- Biemond, February, 1999.
ISSN: 2231-5381
Page 165

Image Restoration Based On Deconvolution by Richardson Lucy Algorithm

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Image Restoration Based On Deconvolution by Richardson Lucy Algorithm

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Image Restoration Based On Deconvolution by Richardson Lucy Algorithm

Uploaded by

Copyright:

Available Formats

International Journal of Engineering Trends and Technology (IJETT) Volume 14 Number 4 Aug 2014

Image Restoration Based On Deconvolution by

PG student, Dept. of Electronics and Communication Engg.

Abstract This article presents the performance analysis of

image restoration. Often the blur kernel is assumed

G (u) = F(u)H(u) +N(u)

are the Fourier transform of g, f, h, .

Keywords Inverse filter, Wiener filter, Lucy- Richardson, MSE

Fig. 1 Image Blurring model

Image deconvolution methods are used to

3. INVERSE DECONVOLUTION METHOD

But if noise gets added to the degraded image

The above equation shows that direct

Fig. 2 Inverse Deconvolution

4. WEINER DECONVOLUTION METHOD

Where E{.} is the expected value of the argument.

The reason for the popularity of the

Where is the convolution operation. F= is the

Non linear iterative technique is better than the

6. RESULTS AND ANALYSIS

Table 2 Results for the peppers image

image affected by blurring

image affected by blurring

image restore by weiner filtering

image restore by Inverse filtering

Fig.4 Results of cameraman.tif (a) original image (b) blurred image

PSNR within the very less iterations for the blind

J. G. Walker, D. A. Fish, A. M. Brinicombe, and E. R. Pike, Blind

Arijit Dutta, Aurindam Dhar, Kaustav Nandy, Project report on Image

G. R. Ayers and J. C. Dainty, Iterative blind deconvolution method and

J. R. einup, Phase retrieval algorithms: a comparison, Appl. Opt.

B. L. K. Davey, R. G. Lane, and R. H. T. Bates, Blind deconvolution of

W. H. Richardson, Bayesian-based iterative method of image

L. B. Lucy, An iterative technique for the rectification of observed

[8] Ramesh Neelamani, Thesis report on Inverse problems in image

You might also like