International Journal of Engineering and Technical Research (IJETR)

ISSN: 2321-0869, Volume-3, Issue-1, January 2015

Motion blur remover using total variation constrained

least squares method with user configurable PSF and
blind PSF estimation
Arpita Agrawal, Vinit Agrawal

blurs can be much more complicated than this simple case.

Abstract removal of blur from the blurred images, which is For example, the blur can be space-variant, nonlinear, local
commonly preliminary step for de-blurring process, has become and multiple.
one of the most growing research areas of these days. Now days
scholars have proposed new methods or improved blur detection We can remove blur by applying these approaches:
algorithms which are based on blind de-convolution, point
spread function (PSF) and many more. Motion blur retains some
information about motion, based on which motion may be
(i) At first we have to estimate that how the image is blurred
recovered from blurred images. This is a difficult problem, as (the blur kernel or the point spread function)
the situations of motion blur can be quite complicated, such as (ii) After it we have to restore the natural looking image by
they may be space variant, nonlinear, and local. A motion blur is using de-convolution.
characterized by its point spread function (PSF) whose
parameters are closely related to the motion. The simplest Blur detection method can be applied as initial stage for
motion blur is the space-invariant linear motion blur, which has de-blurring when the machine vision of manufacturing line is
been studied extensively. In practice, however, as motion can be out of focus or due to rapid movement of the inspected
quite complex, motion blurs can be much more complicated than
product. Some of the restrictions are there in the blur
this simple case. For example, the blur can be space-variant,
nonlinear, local and multiple. The purpose of this paper is to detection technique thats why many researchers research so
extend the work area of already work done. many algorithms to detect more blur from the image [2].
Some of them are belongs to variant methods, point spread
Index Terms Blur detection, blind de-convolution,point functions and de-convolution methods. Based on analysis on
spread function. the past researches on the blur detection methods, blur
detection can be classified in to some categories, which are:
I. INTRODUCTION i) De-blurring using Gaussian method
De-blurring from a single image has been an ill-posed and ii) De-blurring using blind de-convolution method
challenging problem due to the large number of unknowns in iii) Using motion de-blurring method
the estimation process. The unknowns are the type of blur, the
extent of blur and the noise, which degrade the image further. For linear and shift-invariant motion-blurring, the blurred
A great amount of literature exist on the subject motion blur image g(x, y) may be written as the convolution of the scene
remover, this technique is discussed regularly in the various f(x, y) and the point spread function (PSF) h(x, y):
conferences and publications on computer science. In
addition to this a growing number of special issues and g(x, y) =
conferences are dedicated to remove the blur from the images (1)
which are taken and slightly blurred due to camera shake or where n (x, y) represents additive noise. A Fourier trans- form
low light condition of a scene. The articles are selected for of Eq. (1) gives
this issue present new approaches and ideas for the detection
of blur. Blur is the degradation of sharpness and contrast of an G(u, v) = F(u, v)H(u, v) + N(u, v), (2)
image which cause loss of high frequency [1]. For the
de-blurring process we generally apply some algorithm which where G(u, v), F(u, v), H(u, v), and N(u, v) are respectively
estimate the blur and then recover the image into its possible the transforms of g(x, y), f(x, y), h(x, y), and n(x, y). When the
and probable condition. Several methods are used to recover noise term is negligible, a restoration of f(x, y) may be
the blur from the image and some method are used which are obtained by an inverse transform of G(u, v)/H(u, v). This
sometimes hard-coded and sometime user configurable. In restoration technique is often called the inverse filter.
which we use a function called point spread function (PSF).
A motion blur is characterized by its point spread function II. RELATED WORK
(PSF) in which the parameters of PSF are closely related to
the motion. The simplest motion blur is the space-invariant Restoration of blurred images is a vital problem especially in
linear motion blur, which has been studied extensively. In tracking and identification of criminals. The available image
practice, however, as motion can be quite complex, motion can be used to identify a human face or a moving vehicles
number plate taken in hit and run situation or in a bomb blast
site. To restore a blurred image successfully, blurring
Manuscript received January 23, 2015.
Arpita Agrawal, M.Tech Student, Gyan Vihar University, Jaipur, India
function needs to be estimated accurately. Blurring function
Vinit Agrawal, Asst.Professor Gyan Vihar University, Jaipur, India is referred to as Point Spread Function (PSF) which is the

296 www.erpublication.org
Motion blur remover using total variation constrained least squares method with user configurable PSF and blind PSF
estimation

response of an imaging system to a point source or it can 3.2 Motion blur remover using motion de-blurring
be said as the impulse response of a focused optical system method:
[1]. Existing system used a single hardcoded method to
de-blur a natural image. Technologies used in this method are To remove blur a technique is used in which horizontal length
more common but the mathematical approaches are and an angle theta is used at which the image is moved.
sometimes changed with some more configurations. Motion blur retains some information about motion, based on
Technologies which are used in exiting system are motion which motion may be recovered from blurred images. This is
blue remover, Gaussian blur remover and blind estimation a difcult problem, as the situations of motion blur can be
PSF. Along with these techniques only a single technique is quite complicated, such as they may be space- variant,
used which is generally hard coded and not permitted to nonlinear, and local.
change by user.
3.3 Motion Blur remover using blind Point Spread
2.1 Challenges: Function:
The major challenge in de-blurring a natural image is the
determination of the unknown parameters like type of blur, This is the technique in which blur is removed using a PSF
extend of blur (PSF) and the approximation of noise. The which is a function in which a point spread function size is
number of unknowns is more than the number of known described and with the use of this size de-blurring is done. Its
parameters making the problem ill posed. Even minor a unique approach for removing the blur in which we define a
reduction in accuracies in PSF estimation leads to degradation function named point spread function. In which a function
of image quality while de-blurring. describes the imaging system to a point source or a point
object. So with the help of this PSF motion blur can be
III. METHODOLOGY removed.
Blind single-image de-blurring is a method used to restore a IV. PROPOSED WORK
sharp image from an image blurred by camera shakes or
object motion under low light conditions. Generally blur This paper work started with the approach of estimating the
occurred due to motion or low light conditions. To solve this blur parameters using the algorithms discussed below. The
problem many researchers do necessary researches and work algorithm aims at determining the Blur parameters such as
using blurred images and its mathematic approaches. From length of blur in pixels and the angle of blur in degrees. This
the various challenges of photography, one is motion blur. algorithm used for estimating the blur parameters,
Blur problems are generated from relative motion between a de-blurring of the image using the estimated parameters, its
camera and a picture during exposure. While blur can be limitations and the proposed Efficient De-blurring Algorithm
reduced by using a shorter exposure, this comes at an for natural images.
unavoidable trade-off with increased noise. So blur can be
removed computationally. This is the main motive of this V. RESULTS
research. Now the question is that how to remove blur.

We can remove blur by applying these approaches:

(i) At first we have to estimate that how the image is
blurred (the blur kernel or the point spread function)
(ii) After it we have to restore the natural looking
image by using de-convolution.

In the complete procedure of this de-blurring the main

challenge is to estimate the exact figure of natural blurred
images. Another difficulty is that algorithm needs to
distinguish the correct imageblur pair from incorrect ones
that can also adequately explain the blurred image.
De-convolution is also difcult because the algorithm needs
to restore high frequency image contents attenuated by blur.
In this dissertation, we address a few aspects of these
Fig: 1 A restoration obtained with measures of de-blurring
challenges.
using motion constraints.
3.1 Motion blur remover using Gaussian filter:
According to these results we deduce that when restoring a
real-world motion-blurred image, some exploratory
To remove blur a technique is used in which a filter is used
restorations may be carried out using assumed-PSFs, in order
which is named as Gaussian filter. In this we require the size
to determine the shape and blur extent of the true- PSF, guided
of this filter in numeric form and a standard deviation sigma
by the characteristics of the exploratory restorations obtained.
value with which a mathematical calculation is done. To
The estimated true-PSF may then be used for the final
remove the blur from an image we use this technique
restoration. On the basis of the results of the
according to which if there is any movement in the filter size
of any image then this Gaussian filter will work.

297 www.erpublication.org
 International Journal of Engineering and Technical Research (IJETR)
ISSN: 2321-0869, Volume-3, Issue-1, January 2015
On the basis of the above observations, we propose the
following strategy for the exploratory restoration of images In this section we apply the restoration procedure to the major
which have been motion-blurred with an unknown PSF: three techniques of de-blurring. Blurring was caused by
panning the camera when the shutter was being released. The
1. Use a forward-ramp for the first trial restoration. images were digitized with a monochrome video camera with
2. If the restoration from (1) shows a smeared-out scene back illumination.
terminated in a strong, clear ghost, use a reverse- ramp for a
second trial restoration.
3. If the restoration from (1) or (2) shows a ghost, we may
assume a square-pulse or trapezoid true-PSF. Mea- sure the
displacement of the ghost from the true scene to obtain the
blur extent of the true-PSF. Using a square pulse with this
estimated blur extent, another restoration is carried out.
4. If the restoration from (3) shows ringing, try a trape- zoid
PSF with the same estimated blur extent. Use a forward- or a
reverse-trapezoid PSF depending on whether a forward- or a
reverse-ramp was found more appropriate in steps (1) and
(2).For more demanding applications, we may use the above
procedure to establish the overall structure of the true-PSF
and then experiment with fine variations in PSF shapes to
obtain a really high quality restoration. An advantage of the Fig: 4 A motion-blurred photographs taken by panning the
above procedure is that the true-PSFs may be determined to camera.
reasonable accuracy even if they are not spatially invariant.
In Fig. 4, the text is badly blurred, but the holes and pins are
clearly visible. Figure 5 show the image restored using motion
de-blurred process. Pin numbers and texts are clearly visible.

Fig: 2 a restoration obtained with measures of de-blurring

using the size of PSF Measure
Fig: 5 restored image with motion de-blurring process.

Fig: 3 A Restoration obtained with measures of de-blurring Fig: 6 de- blurred photograph of motion blurred photograph
using the Gaussian de-blurring process. taken by camera.
5.1 SOME REAL-WORLD MOTION-BLURRED IMAGES
RESTORED:

298 www.erpublication.org
Motion blur remover using total variation constrained least squares method with user configurable PSF and blind PSF
estimation

VI. SUMMARY AND CONCLUSION

A. Summary
When we took a picture sometimes it shakes due to some
motion or low light. Due to this blurring of image occurs. To
address this problem, some researches do their efficient work
on de-blurring dynamic scenes and prepare an approach that
commonly requires accurate motion segmentation. To
explore the existing systems approach we differentiate it with
the previously done work and the new approach which we are
doing. Till yet the de-blurring is done by applying several
methods of de-blurring but single approach in single time. We
are applying a hardcoded system according to which user can
configure the approach for de-blurring.
Estimating the PSF of a real-world motion-blurred image is an
essential step in the restoration process. This paper addresses
the problem of blind motion de-convolution without assuming
that the entire image undergone the same blur. In this we
apply the procedure of de-blurring with several techniques.
We apply the
B. Conclusion:
The research includes several concepts of image processing
and tools of MATLAB to de-blur the blurred image which is
blurred by camera shake. In this research many techniques of
de-blurring is used which can be of many procedures. The
operated system works in a very fine way and work on the
estimation of PSF which is point spread function. Our
approach takes advantage of both images to produce a high
quality reconstructed image. By formulating the image
de-blurring problem using two images, we have developed an
iterative de-convolution algorithm which can estimate a very
good initial kernel and signicantly reduce de-convolution
artifacts. No special hardware is required.

ACKNOWLEDGMENT

It is my pleasure to express my hearty gratitude to my teacher

and guide, Mr. Vinit Agrawal, (Assistant Professor), Faculty
of Engineering, Department of Computer Science &
Engineering, Suresh Gyan Vihar University, Jaipur, who
inspired and initiated me to prepare this research despite his
busy academic schedule. He has always been kind enough to
spare his valuable time and thought in giving necessary
guidance. His rich experience as an academician helped me in
understanding this topic clearly. Finally, words are
insufficient to express my profound sense of gratitude to my
parents whose motivation and blessing always provides me
with great strength.

REFERENCES

[1]. M. Cannon, Blind deconvolution of spatially invariant image blurs with

phase, IEEE Trans. ASSP 24, 1976, 58-63.
[2] Wikipedia, Point spread function --- Wikipedia
[3] N. Joshi, R. Szeliski, and D. J. Kriegman, "PSF estimation using sharp
edge prediction," in Proc. IEEE Conf. Computer Vision and Pattern
Recognition CVPR 2008, 2008, pp. 1-8.
[4] Qi Shan, Jiaya Jia, and Aseem Agarwala, "High-quality motion
deblurring from a single image," in ACM SIGGRAPH 2008 papers,
2008, pp. 73:1--73:10.

299 www.erpublication.org