Image Deblurring
Image Deblurring
Image Deblurring
International Journal of Advances in Computer Science and Technology (IJACST), Vol.3 , No.12, Pages : 01-08
Special Issue of ICCEeT 2014 - Held on 22nd December 2014, Dubai
INTRODUCTION
Images are widely used in many kinds of applications such
as everyday photography, monitoring, medical imaging,
astronomy, microscopy, and remote sensing. Digital images
are composed of picture elements or pixels that are organized
in a grid. Each pixel contains an intensity value which
determines the tone at a specific point. Unfortunately, all
captured images end up more or less blurry. The motion of
objects or the vibration of the sensor (camera) when pressing
the shutter causes the image to be blurred. There are many
factors that cause the blurring or degradation of the digital
image, such as movement during the capture process, using
long exposure times, using wide angle lens, etc.[2]. However,
there are two main causes for motion blur: (i) the image is
blurred by the camera vibration which causes all pixels in the
image to be affected, and (ii) the image is blurred by object
motion which causes a specific region to be blurred. Image
blur usually devastate the images, and practically it is hard to
avoid it because there is a lot of interference in the
environment. Image deblurring is the process of applying and
solving mathematical models to recover the original (sharp)
image. Recently, image deblurring and restoration become
the main and the most important sub field of digital image
processing. Image deblurring is used to make images sharp
and useful. There are many applications for image deblurring
such as:
(Eq. 1)
ISSN 2320-2602
International Journal of Advances in Computer Science and Technology (IJACST), Vol.3 , No.12, Pages : 01-08
Special Issue of ICCEeT 2014 - Held on 22nd December 2014, Dubai
think that zooming is one type of blurring. Actually, if an
image is zoomed using pixel replication and large zooming
factor, the resulting image is a blurred image. however, the
difference between zooming and blurring is that the number
of pixels in the zoomed image is larger than the number of
pixels in the original image, but the number of pixels in the
original image and blurred one is the same. Blurring can be
thought of as applying some filter to an image. There are
many kind of filters that can be used to perform blurring, such
as average filter, weighted average filter, Gaussian filter and
motion filter. The average filter has three properties: it is odd
ordered, the sum of all its elements equal 1, and all its
elements are the same. The amount of the blurring is
increased when the size of the kernel is increased. That is
because a large number of pixels are included and one
smooth transition is defined. The weighted average filter
gives more weight to the center value. As a result, the
contribution of the center pixel becomes more than the rest of
the pixels. The Gaussian blur uses a Gaussian function to
calculate the transformation that will be applied to each pixel
in the image. It is low-bass pass filter which is always used to
reduce noise and enhance image structure at different scales.
In Gaussian blur, the pixel weights are decreased from kernel
centre to edges according to a bell-shaped curve. The
Gaussian filter blends a specific number of pixels
incrementally, following a bell-shaped curve. The blurring is
dense in the centre and feathers at the edge[3]. It can be used
to give more control over the Blur effect because the
Gaussian blur depends on the Size and Alfa. Motion blur is
the apparent streaking of rapidly moving objects in a still
image[5]. It occurs when the image being recorded is
changed during the recording either due to rapid movement
or long exposure. The basic reason for all motion blur is the
relative motion between the recording device and the scene in
the form of translation, rotation, sudden change of scale, or a
combinations of them [2]. The Motion Blur is a filter that
adds a blur in a specific direction, so the image appears to be
moving. The motion is controlled by an angle (0 to 360
degree) or direction (-90 to 90) and by distance or intensity in
pixels. There are many other types of blur types such as
out-of-focus blur. If a camera is used to convert a 3D real
scene into a 2D image, some parts of the scene are in focus
while other parts are not. If the aperture of the camera is
circular, the image of any point source is a small disk, known
as the circle of confusion (COC)[6]. The degree of defocus
(diameter of the COC) depends on several factors including:
the focal length, the aperture number of the lens, and the
distance between the camera and the scene.
(Eq. 6)
(Eq. 7)
DEBLURRING TECHNIQUES
(Eq. 8)
ISSN 2320-2602
International Journal of Advances in Computer Science and Technology (IJACST), Vol.3 , No.12, Pages : 01-08
Special Issue of ICCEeT 2014 - Held on 22nd December 2014, Dubai
inverse of the blurring matrix A can be expressed as shown in
equation 9.
ITERATIVE METHODS
(Eq. 9)
(Eq. 10)
1
(Eq. 11)
(Eq. 12)
(Eq. 13)
(Eq. 14)
=
=
( )
(Eq. 15)
( )
(Eq. 16)
ISSN 2320-2602
International Journal of Advances in Computer Science and Technology (IJACST), Vol.3 , No.12, Pages : 01-08
Special Issue of ICCEeT 2014 - Held on 22nd December 2014, Dubai
WIENER FILTERING
(Eq. 17)
SPARSE REPRESENTATION
Most of the blind image deblurring techniques either only
remove simple motion blurring, or need user interactions to
handle more complex cases. Sparse decomposition is a
process that estimates a sparse multi-dimensional vector that
satisfies a set of equations in a linear system given
high-dimensional observed data and a design matrix. It is
used in a wide range of applications such as image deblurring
and restoration. Sparse representation succeeds due to the
development of l1-norm optimization techniques, and also
because natural images are intrinsically sparse in some
domain. Some of the techniques that use sparse
approximation for blind image deblurring can be found on
[55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69]. J.
Yang et al. [54] presented an approach for image
super-resolution using sparse signal representation. From
image statistics, they stated that image patches can be
well-represented as a sparse linear combination of elements
from an appropriately chosen over-complete dictionary. So
they seek a sparse representation for each patch of the
low-resolution input, and then use the coefcients of this
representation to generate the high-resolution output. W.
Dong et al. [52] introduced two adaptive regularization terms
into the sparse representation framework. The first term is a
set of autoregressive (AR) models that are learned from a
specific dataset. The best fitted AR model is adaptively
selected to regularize the image local structures. The second
term is the image non-local self-similarity. Experimental
results on image deblurring and super-resolution shows that
the usage of adaptive sparse domain selection and adaptive
regularization, achieves an improved results in terms of
PSNR and visual perception. J.-F. Cai et al. [53] presented an
approach to remove motion blurring from a single image by
formulating the blind blurring as a new joint optimization
problem, which simultaneously maximizes the sparsity of the
blur kernel and the sparsity of the clear image under certain
suitable redundant tight frame systems (curvelet system for
kernels and framelet system for images). The advantage of
this techniques is that it can recover high-quality images from
given blurred images without any prior knowledge of the blur
kernel.
(Eq. 18)
( , ) ( , )
| ( , )| ( , ) + ( , )
(Eq. 19)
(Eq. 20)
ISSN 2320-2602
PSNR
International Journal of Advances in Computer Science and Technology (IJACST), Vol.3 , No.12, Pages : 01-08
Special Issue of ICCEeT 2014 - Held on 22nd December 2014, Dubai
40
35
30
25
20
15
10
5
0
Y. Tai et S. Prasad
al[9]
et al[11]
Series1
30
30
O.
Whyte
et al[79]
A.-P.
Yang et
al[32]
30
20
Z.
J.
J. L.
Hongyin L. Yuan
P. Singh S. kumar W. Dong J.-F. Cai
Biemond Herring
g et
et al[73]
et al[38] et al[40] et al[52] et al[53]
et al[75] et al[77]
al[34]
19
30
29
22
20
26
35
33
SSIM
0.8
0.6
0.4
0.2
0
Y. Tai et
al[9]
Series1
0.93
S.
O.
Prasad Whyte
et al[11] et al[79]
0.95
0.9
A.-P.
Yang et
al[32]
0.76
Z.
J.
J. L.
Hongyin L. Yuan
P. Singh S. kumar W. Dong J.-F. Cai
Biemond Herring
et al[73]
et al[38] et al[40] et al[52] et al[53]
g et
et al[75] et al[77]
al[34]
0.77
0.93
0.88
0.69
0.69
0.72
0.97
0.98
OTHER APPROACHES
As we discussed before, non-blind deconvolution methods
have been widely used for image deblurring. These
techniques can handle only the linear blur model where the
blurred image is generated by a linear convolution of the
latent image and the blur kernel. However, this often fails
because there are several types of outliers that should be
handled. S. Cho et al. [70] proposed a novel blur model that
explicitly takes into account several types of outliers, and
build a robust non-blind deconvolution method that reduces
the visual artifacts caused by outliers. They tested the
proposed system on synthetic and real-world examples. It
was shown that if these outliers are not handled, they cause
severe ringing artifacts even if the kernel is estimated
accurately. So they analyze a few common types of outliers
that cause other deblurring techniques to fail, such as pixel
saturation and non-Gaussian noise. H. Yin and I. Hussain[72]
proposed a blind deconvolution method for image deblurring
based on ICA measure as well as a simple genetic algorithm.
S. El-Regaily, et al. [71] proposed an algorithm for image
deblurring using Genetic Algorithms, by finding proper
parameters and goal function.
O. Whyte et al. [79] propose a new parametrized geometric
model of the blurring process in terms of the rotational
motion of the camera during exposure. The model can
capture non-uniform blur in an image due to camera shake
using a single global descriptor, and can be substituted into
existing deblurring algorithms with only small modifications.
L Yuan et al. [73] proposed an approach that uses the blurred
image and an assistance noisy image to estimate an accurate
ISSN 2320-2602
International Journal of Advances in Computer Science and Technology (IJACST), Vol.3 , No.12, Pages : 01-08
Special Issue of ICCEeT 2014 - Held on 22nd December 2014, Dubai
image in dB (decibels). PSNR between two images
(deblurred image and original image) measures how far the
two images are equal.
1
= 10
(Eq. 21)
)
(Eq. 22)
(2
(
+ )(2
+ )
+ )( +
+ )
[16] H. L. Yang, Y.-H. Chiao, P.-H Huang, and S.-H. Lai. "Blind image
deblurring with modified richardson-lucy deconvolution for ringing
artifact suppression." In Advances in Image and Video Technology,
pp. 240-251. Springer Berlin Heidelberg, 2012.
[17] M. K.Khan, S. Morigi, L. Reichel, and F. Sgallari. "Iterative methods
of Richardson-Lucy-type for image deblurring." Numerical
Mathematics: Theory, Methods and Applications 12,2003.
(Eq. 23)
[18] J. L. Wu, C.-F. Chang, and C.-S. Chen. "An adaptive richardson-lucy
algorithm for single image deblurring using local extrema
filtering." Journal of applied science and engineering 16, no. 3, 2013.
[19] Y. W. Tai, P. Tan, L. Gao, and M.S.
Richardson-Lucy deblurring for scenes under projective
motion path, Technical report, KAIST, 2009.
[20] J. J. Ding, W.-D. Chang, Y. Chen, S.-W Fu, C.-W Chang, and C.-C.
Chang. "Image deblurring using a pyramid-based Richardson-Lucy
algorithm." In Digital Signal Processing (DSP), 2014 19th
International Conference on, PP. 204-209. IEEE, 2014.
[21] Y. Chu, S. Li, and M. Wang. "A Modified Richardson-Lucy Method
for Space-Variant Motion Deblurring." In Computer Science &
Service System (CSSS), International Conference on, pp. 1884-1887.
IEEE, 2012.
REFERENCES
[1]
[2]
[3]
[4]
[5]
Brown,
[23] H.-L. Yang, P.-H. Huang, and S.-H. Lai. "A novel gradient attenuation
RichardsonLucy algorithm for image motion deblurring." Signal
Processing 103, PP. 399-414, 2014.
[24]
[6]
[7]
[8]
[9]
[32] A.-P. Yang, Z.-X. Hou, and C.-Y. Wang. "Image deblurring based on
wavelet and neural network." In Wavelet Analysis and Pattern
Recognition, 2007. ICWAPR'07. International Conference on, vol. 2,
pp. 647-651. IEEE, 2007.
ISSN 2320-2602
International Journal of Advances in Computer Science and Technology (IJACST), Vol.3 , No.12, Pages : 01-08
Special Issue of ICCEeT 2014 - Held on 22nd December 2014, Dubai
[33] N. Kumar, M. Shrivastava, and N. Rahim. "Image Deblurring Using a
Neural Network Approach", International Journal of Engineering and
Innovative Technology (IJEIT), Vol. 2, No. 3,September 2012
[53] J.-F. Cai, H.Ji, C. Liu, and Z.Shen. "Blind motion deblurring from a
single image using sparse approximation." In Computer Vision and
Pattern Recognition, 2009. CVPR 2009. IEEE Conference on, pp.
104-111. IEEE, 2009.
[37] A.-P. Hou, Y.Z.-Xin, and C.-Y. Wang. "image deblurring based on
wavelet and neural network", Wavelet Analysis and Pattern
Recognition, 2007. ICWAPR'07. International Conference on, 2007.
[59] L. Ma, J. Yu, and T.Zeng. "Sparse Representation Prior and Total
Variation--Based Image Deblurring under Impulse Noise." SIAM
Journal on Imaging Sciences 6, no. 4, PP. 2258-2284, 2013.
[60] H. Huang, and N. Xiao. "Image deblurring based on sparse model with
dictionary learning [J]." Journal of Information and Computational
Science10, no. 1, PP. 129-137, 2013.
[61] W. Dong, G. Shi, and X. Li. "Image deblurring with low-rank
approximation structured sparse representation." In Signal &
Information Processing Association Annual Summit and Conference
(APSIPA ASC), Asia-Pacific. 2012.
[44] Y. Zheng, "A Fast Image Deblurring Algorithm Using the Wiener
Filter and the Hartley Transform." In Review of Progress in
Quantitative Nondestructive Evaluation, pp. 735-742. Springer US,
1989.
[64] M. Yin, J. Gao, D. Tien, and S. Cai. "Blind image deblurring via
coupled sparse representation." Journal of Visual Communication and
Image Representation 25, no. 5, PP. 814-821, 2014.
[72] H. Yin, and I. Hussain. "ICA and genetic algorithms for blind signal
and image deconvolution and deblurring." In Intelligent Data
Engineering and Automated LearningIDEAL 2006, pp. 595-603.
Springer Berlin Heidelberg, 2006.
ISSN 2320-2602
International Journal of Advances in Computer Science and Technology (IJACST), Vol.3 , No.12, Pages : 01-08
Special Issue of ICCEeT 2014 - Held on 22nd December 2014, Dubai
[73] L. Yuan, J. Sun, L. Quan, and H. Shum, Image deblurring with
blurred/noisy image pairs, ACM Trans. Graph., vol. 26, no. 3, pp.
111,2007.
[74] A. Gupta, N. Joshi, C. Lawrence Zitnick, M. Cohen, and B. Curless,
Single image deblurring using motion density functions, in Proc
ECCV, PP. 171184, 2010.
[75] J. Biemond, R. L. Lagendijk, and R. M. Mersereau. "Iterative methods
for image deblurring." Proceedings of the IEEE 78, no. 5, PP. 856-883,
1990.
[76] A. Beck, and Marc Teboulle. "A fast iterative shrinkage-thresholding
algorithm with application to wavelet-based image deblurring."
In Acoustics, Speech and Signal Processing, 2009. ICASSP 2009.
IEEE International Conference on, pp. 693-696. IEEE, 2009.
[77] J. L. Herring,"The Effects of Projection on Iterative Methods in Image
Deblurring." PhD diss., Emory University, 2010.
[78] J. G. Nagy, K. Palmer, and L.Perrone. "Iterative methods for image
deblurring: a Matlab object-oriented approach." Numerical
Algorithms 36, no. 1, PP. 73-93, 2004.
[79] O. Whyte, J.Sivic, A.Zisserman, and Jean Ponce. "Non-uniform
deblurring for shaken images." International journal of computer
vision 98, no. 2, PP. 168-186, 2012.