Image restoration

MOTION BLUR IMAGE RESRORATION USING ALTERNATING DIRECTION BALANCED REGULARIZATION FILTER

A DISSERTATION

PresentedIn partial fulfillment of the requirement for the award of degree of

MASTER OF TECHNOLOGYIN

COMPUTER SCIENCE & ENGINEERINGSubmitted by

Manoj Kumar Rajput(0901CS13MT06)

 Under the supervision ofRajeev Kumar Singh(Assistant Professor)

Department of Computer Science & Engineering and Information TechnologyMadhav Institute of Technology & Science, Gwalior (MP) - 474005

  Session 2013-2015

ContentsIntroductionLiterature ReviewProposed MethodologyExperimental AnalysisConclusions and Future ScopeReferencesPublication

Introduction Image restoration [1] is a technique used to recover image from degraded image. image may be

distored due to blur and noise; blur can occur due to atmospheric turbulence, motion of objects and camera miss-focus.

The degradation model of imagine system can be define as ,

Where, is the degraded image , in which denotes the two-dimensional linear convolution operation, is the original image and is point spread function is the additive noise.

Motion blur is due to relative motion between the recording device and the scene. When an object or the camera is moved during light exposure, a motion – blurred image is produced.

When the scene to be recorded translates relative to the camera at a constant velocity under an angle of radians with the horizontal axis during the exposure interval [3], the distortion is one-dimensional.

defining the length of motion by

kKk nhIg

kg kI h

kn

relativev osuretexp,0

osurerelative tvL exp

Continue… Atmospheric motion blur image restoration can be done in two stages: first stage is

restoration and second stage is post processing of image. The effects to atmospheric turbulence can be measured by calculating the scintillation

index [2]. This index is related to the mean and standard deviation of the intensity distribution of image.

The atmospheric turbulence can be categorized in two categories: rigid and non-rigid bodies .

In rigid bodies restoration, we take single frame image which is degraded through turbulence and restoration techniques will be applied in degraded image, which is known as single image deconvolution [4].

The point spread function[11] of atmospheric turbulence motion blur image can be described below,

The estimation of point spread function of atmospheric turbulence image is very challenging without knowing prior knowledge of clean image[5] . So, it is difficult to restore image using point spread function (PSF) .

Example of images

a) Original Lena image b) Motion blur Lena image

c) Restored Lena image

Figure 1. Effect of motion blur on image due to atmospheric turbulence

Literature Review Xie Kai and Li Tong [7] have proposed the Herr wavelet transform (HWT) in discriminating

different types of edges as well as recovering sharpness from the blurred version, and then determines whether an image is blurred or not and up to what extent if it is blurred. Two different schemes are proposed to estimate the motion blur parameters. Two dimensional Gabor filter has been used to calculate the direction of the blur. Radial basis function neural network (RBFNN) has been utilized to find the length of the blur. Subsequently, Wiener filter has been used to restore the images. Noise robustness of the proposed scheme is tested with different noise strengths. The blur parameter estimation problem is modeled as a pattern classification problem and is solved using support vector machine (SVM).

R.Dash, P. K. SA, and B. Majhi [6] have introduced approach to estimate the motion blur parameters using Gabor filter for blur direction and radial basis function for blur length with sum of Fourier coefficients as features. Restoration attempts to recover an image by modeling the degradation function and applying the inverse process. Motion blur is a common type of degradation which is caused by the relative motion between an object and camera. Motion blur can be modeled by a point spread function consists of two parameters angle and length. Accurate estimation of these parameters is required in case of blind restoration of motion blurred images. This paper compares different approaches to estimate the parameters of a motion blur namely direction and length directly from the observed image with and without the influence of Gaussian noise. These estimated motion blur parameters can then be used in a standard non-blind de-convolution algorithm.

Continue… Li and Simske [8] have introduced for atmospheric turbulence blurred images. They

have used the concept that kurtosis of an image increases with extent of blurring. Phase structure has been utilized to analyze the impact of blurring on kurtosis. Blur parameter is estimated after setting the search space on a trial and error basis. For each of the estimated parameter, the image is de-blurred using a classical image restoration technique.

Haiyong Liao, Fang Li, and Michael K [9] have introduced a fast image restoration method which selects the regularization parameter automatically to restore noisy blurred images. The method exploits the generalized cross validation technique to determine the amount regularization used in each restoration step. The regularization parameter is updated each iteration, which increases the closeness of the restored image towards the true image.

F. Krahmer, Y. Lin, B. McAdoo, K. Ott, J. Wang, D. Widemann, and B. Wohlberg [10] have focused on Radon transform for searching characteristics of motion blur in capstan analysis. This report discusses methods for estimating linear motion blur. The blurred image is modeled as a convolution between the original image and an unknown point-spread

Proposed MethodologyThe following steps are performed given below:Step-1. Take input image.Step-2. Apply length and angle in an image.Step-3. Get motion blurred image and also add Gaussian noise with 0.001 density. Step-4. Apply Gabor filters to estimate angle.Step-5. Apply Radial basis function (RBF) to estimate length.Step-6. Calculate point spread function (PSF) with the help of estimated parameter.Step-7. The image is restored using alternating direction balanced regularization (ADBR) Filter.Step-8. The calculate peak signal to noise ratio (PSNR), Mean square error (MSE) and Elapsed time.

Continue…

Input image

Apply length and angle in imageAnd add Gaussian noise with

0.001 density.

Obtain Motion blurred image

Estimate angle using Gabor filter

Estimate length using Radial basis function (RBF)

Calculate Point spread function (PSF) with the help of estimated

parameter

Apply Alternating direction balanced regularization (ADBR)

filter for restored image

Calculate PSNR, MSE and elapsed time

End

Obtain Restored image

Figure 2. Flow Chart and GUI of Proposed work

Start

Experimental Analysis The analysis of result obtain by proposed methodology is done by comparing the result of

other existing methods. on the basis of two parameter, MSE and PSNR. In the restoration, the imperceptibility or the quality of the image is measured by using peak

signal to noise ratio (PSNR)[11]. PSNR is used to calculate the similarity in the original image and the motion blurred image. the PSNR is calculated by using two images one is the original image and other is the restored image. the value of the PSNR is always greater the 40. the basic formula of PSNR is given below:

Where , MSE the mean square error between two image.

MEAN SQUARE ERROR (MSE)

The MSE is used to measures average squared disparities between the original image and the restored image[12]. the formula of MSE is given below. In the equation and respectively

MSEPSNR

2

10255log10

Continue…

Where, and are the gray scale values of original and restored image

is the size of image. Firstly the MSE (mean square error) will be calculated then the PSNR ( peak signal

to nose ratio) value is calculated. There fore , higher value of PSNR denotes less distortion

1

0

1

0

2,

^,

1 M

i

N

jjiji xx

NMMSE

jix ,^x NM

Apply method in different types of Image

(a) original lenna image (b) degrade image (c) restored image

(a) original Barbara image (b) degrade image (c) restored image

Figure 3. Lena image

Figure 4. Barbara image

Continue…

(a) original cameraman image (b) degrade image (c) restored image

(a) original Boat image (b) degrade image (c) restored image

Figure 5. Cameraman image

Figure 6. Boat image

Continue…

(a) original house image (b) degrade image (c) restored image

(a) original pappers image (b) degrade image (c) restored image

Figure 7. House image

Figure 8. Pappers image

Continue…

(a) original stick image (b) degrade image (c) restored image

(a) original group image (b) degrade image (c) restored image

Figure 9. Stick image

Figure 10. Group image

Continue… MSE (Mean Square Error) value are shown in table 1.

Table 1: Mean Square Error (MSE) value between existing and proposed method

* Motion blur parameters estimation for image restoration

# Motion Blur Image Restoration Using Alternating Direction Balanced Regularization Filter

S.No Image MSE calculated by Base Method

MSE calculated by

Proposed Method

1 Lena 0.0158 0.0090

2 Barbara 0.0087 0.0219

3 Cameraman 0.0157 0.0107

4 Boat 0.0119 0.0093

5 House 0.0149 0.0083

6 Pappers 0.0092 0.0027

7 Stick 0.0073 0.0070

8 Group image 0.0271 0.0047

#*

Continue… Peak Signal to Noise Ratio (PSNR) value are shown in table 2.

S.No Image PSNR Value calculated by Base Method

PSNR value calculate Proposed Method

1 Lena 66.16 68.54 2 Barbara 61.74 64.71 3 Cameraman 66.16 67.82 4 Boat 67.36 68.42 5 House 66.38 68.94 6 Pappers 68.49 73.84 7 Stick 69.50 69.67 8 Group image 63.79 71.32

Table 2: Peak Signal to Noise Ratio (PSNR) value between existing and proposed method.

* #

* Motion blur parameters estimation for image restoration

# Motion Blur Image Restoration Using Alternating Direction Balanced Regularization Filter

Continue…

Figure 11. Comparison between existing Method and Proposed Method MSE on different images.

Continue…

Figure 12. Comparison between Base Method and Proposed Method PSNR on different images.

Conclusion The restoration results in the improved visualization of the image. This work

presented an Alternating Direction Balanced Regularization Method for finding restored image

which is useful for enhancing the peak signal to noise ratio (PSNR) value for that image. In this work, mean square error (MSE) value of an image decreases in such a way that gives optimized and enhanced image. Proposed algorithm takes less execution time as compared to existing methods.

In this study, Gabor filter and radial basis function have been used to estimate blur angle and blur length respectively. Performance analysis has been made on only blurred images as well as noisy blurred images. The proposed scheme estimates the blur parameters close to the true value. Comparative analysis demonstrates the efficiency of the proposed method.

This implicates the robustness of the proposed method. Both standard and real-time images have been included in experiment. It has been observed in all cases the proposed method provides better result. The experimental analysis of this work shows that proposed algorithm gives better and optimized restored image.

Future Scope There is hope to build new methodology which increases peak signal to noise ratio (PSNR)

value of restored image which provides more accurate and efficient results through newly optimization techniques.

References [1] Dalong Li and Steven Simske, “Atmospheric Turbulence Degraded Image by Kurtosis Minimization”, IEEE Geoscience and RemoteSensing Letters, vol.13, pp 63-69, Dec. 2008.

[2] R.E. Hufnagel and N.R. Stanley, “ Modulation transfer function associated with image through turbulence media”, J. opt. Soc. Amer: A, Opt. Image Sci., vol. 54, pp 52-61, 1964.

[3] K. He, J. Sun and X. Tang, “Single image haze removal using dark channel prior”, Proc. of CVPR, vol 15, PP 1956-1963, June 2009.

[4] D. Li, S. Simske , and R.M. Mersereau, “ Blind Image deconvolution using constrained variance maximization”, in proc. Asilomar Conf. Signals, Syst., Comput, vol 08, pp. 1762- 1765, 2004.

[5] Luxin Yan, Mingzhi Jin, Houzhang Fang, Hai Liu, and Tianxu Zhang, “Atmospheric- Turbulence-Degraded Astronomical Image Restoration by Minimizing Second-Order Central moment,” IEEE Geoscience And Remote Sensing Letters, Vol. 9, pp. 672-676, July 2012.

[6] R. Dash, P. K. Sa, and B. Majhi, “RBFN based motion blur parameter estimation”, in Proc. IEEE International Conference on Advanced Computer Control, Singapore, vol 18, PP 327-33, Jan 2009.

[7] Xie Kai and Li Tong, “Arnoldi process based on optimal estimation of the regularisation parameter”, In IEEE International Workshop on Imaging Systems and Techniques, vol 16, PP 340 – 343. 2009.

Continue… [8] Nilanjan Dey, Anamitra Bardhan Roy and Sayantan Dey, “A Novel Approach of Color Image Hiding using RGB Color Planes and DWT ”, International Journal of Computer Applications, vol 6 ,PP 19-22, 2011, [9] Haiyong Liao, Fang Li, and Michael K. Ng, “Selection of regularization parameter in total variation image restoration”, Journal of Optical society of America, vol 16, PP 2311 – 2320, 2009.

[10] F. Krahmer, Y. Lin, B. McAdoo, K. Ott, J. Wang, D. Widemann, and B. Wohlberg, “Blind image de-convolution: Motion blur estimation”, Tech Rep., Univ. Minnesota, vol 6 ,PP 478- 482, 2006.

[11] Jin-Bao Wang, Ning He, Lu-Lu Zhang, and Ke Lu, “Single Image dehazing with a physical model and dark channel prior”, Elsevier, neurocomputing ,vol 9 , pp 312-317,Aug 2014.

[12] G. M. Gluckman, “Kurtosis and the Phase Structure of Images,” in 3rd International Workshop on Statistical and Computational Theories of Vision, Nice, France, October 2003 (in conjunction with ICCV’03), Nice, France, vol 7 , pp12–15, 2003.

[13] K. Gibson and T. Nguyen, “Fast single image fog removal using the adaptive wiener filter,” in Proc. 20th IEEE ICIP, vol 11, pp. 714–718, Sep. 2013.

[14] J.-P. Tarel and N. Hautiere, “Fast visibility restoration from a single color or gray level image,” in Proc. IEEE 12th Int. Conf. Comput. Vis, vol 13, pp. 2201–2208,Oct. 2009.

Continue…[15]Amandeep Kaur, Vinay Chopra, “A Comparative Study and Analysis of Image

RestorationTechniques Using Different Images Formats”, International Journal for Science and Emerging Technologies with Latest Trends, vol 8,PP 7-14,2012.

[16]S. Anna durai and R. Shanmuga lakshmi, “Fundamentals of Digital image Processing", Published by Dorling Kindersley (india) Pvt. Ltd., licensees of Pearson Education in South Asia, vol 14, PP 978-983, 2009.

[17]Y. Yorozu, M. Hirano, K. Oka, and Y. Tagawa, “Electron spectroscopy studies on magneto-optical media and plastic substrate interfaces (Translation Journals style)”, IEEE Transl. J. Magn.Jpn., vol. 2, PP 740–741,Aug. 1987.

[18]Ullah, R. Nawaz, and J. Iqbal, "Single image haze removal using improved dark channel prior." Modelling, Identification & Control (ICMIC), Proceedings of International Conference on. IEEE, PP 154-159, 2013.

[19]G. R. Faulhaber, “Design of service systems with priority reservation”, in Conf. Rec. IEEE Int. Conf. Communications,vol 9, PP 3–8,2005.[20]Neelamani , Choi , and Baraniuk, “Fourier-wavelet regularized de-convolution for ill conditioned

systems”, IEEE Trans. on Signal Processing, vol 14,PP 891-899,2003.

PublicationManoj Kumar Rajput, Rajeev Kumar Singh, “A Review On Image

Restoration Techniques”, International Journal Of Advanced Technology & Engineering Research , Volume 5, Issue 6, pp. 1-7,ISSN: 2250-3536, Nov2015. (published)

Manoj Kumar Rajput, Rajeev Kumar Singh, “Image Restoration of Motion Blur Image using Alternating Direction Balanced Regularization Method”, International Journal Of Communication Systems And Network Technologies (IJCSNT) , ISSN: 2053-6283, 2015. (Accepted)

Thank you