Abstract—A new method, based on the normal shrink and modified version of Katssagelous and Lay, is proposed for multiscale blind image restoration. The method deals with the noise and blur in the images. It is shown that the normal shrink gives the highest S/N (signal to noise ratio) for image denoising process. The multiscale blind image restoration is divided in two sections. The first part of this paper proposes normal shrink for image denoising and the second part of paper proposes modified version of katssagelous and Lay for blur estimation and the combination of both methods to reach a multiscale blind image restoration. Keywords—Multiscale blind image restoration, image denoising, blur estimation. I. INTRODUCTION MAGE are the main sources of information in different field such as medical imaging, astronomy, public security, and satellite imaging. But, due to some reasons, observed images are degraded which are mainly caused by blur and noise. Therefore, image restoration is necessary. Many conventional approaches have been developed to restore the original image require the prior knowledge of blur and some features of noise [7]. Unfortunately in most cases such knowledge is not available and the blind image restoration should be used to restore original image. Blind restoration is a far more complicated problem than simple image restoration. One of the most important tools for blind image restoration is wavelets. Wavelets are mathematical functions that cut up data into different frequency components and then each component is studied with different resolution matched to its scale [2] since Gaussian noise is one of the factor that was caused the image was degraded. Then, removing noise is necessary. For removing noise we use from Normal shrink in the scale of wavelet. In the second part of paper, we use from modified method of Katssagelous and Lay for blur estimation. Finally we use both methods for Multiscale blind image restoration. Several blind restoration algorithms have been proposed in the past. Multiscale Blind image Restoration using a wavelet This work was supported in part by the Islamic Azad University of Bushehr, Iran. Alireza Mallahzadeh is with the Islamic Azad University of Bushehr, Iran (e-mail: Malahzad@yahoo.com). Iman Elyasi is with the Islamic Azad University of Bushehr, Iran (e-mail: iman765765@yahoo.com). Hamid Dehghani is with the Malek-Ashtar University of Technology, Iran (e-mail: hamid_deh@yahoo.com). decomposition was proposed by Sze-Ho in 1996[3]. Spatially adaptive wavelet-based Multiscale image restoration was proposed by Mark.R.banham in 1996[4]. Blind image restoration for MMW radiometer based on wavelet techniques was proposed by Hyuk Park in 2005[1]. Adaptive Wavelet thresholding for image denoising and compression was proposed in 2000 [5] and some surveys can be found in [6]. This paper is structured as follows: in section II the image restoration formulation is reviewed. Section III deales with the wavelet threshold and section IV describes blur estimation. In section V The proposed new method for multiscaled blind image restoration is described. Sections VI and VII describe the experimental results and conclusions respectively. II. FORMULATION The process of image degradation is shown in Fig. 1. Fig 1 Block diagram degraded model Where ) , ( y x f represents an original image and ) , ( y x Y is the degraded image ، ) , ( y x V represents an additive noise (usually, noise is Gaussian noise with zero mean) ، ) , ( y x h is the point spread function of blur. For more details, the reader is referred to [7]: ) , ( ) , ( ) , ( ) , ( y x V y x h y x f y x Y + ⊗ = (1) The expression of the equation in the frequency domain by the Fourier transform is ) , ( ) , ( ) , ( ) , ( v u V v u H v u F v u Y + = (2) In the blind image restoration we should identify PSF and noise as estimated image get similar original image approximately. III. WAVELET THRESHOLD A. Description of Wavelet Threshold Let the original image be } ,..., 1 , ) , ( { N y x y x f = where N is some integer power of 2. The original image is corrupted by additive noise and one observes Multiscale Blind Image Restoration with a New Method Alireza Mallahzadeh, Hamid Dehghani, and Iman Elyasi I World Academy of Science, Engineering and Technology International Journal of Computer and Information Engineering Vol:2, No:5, 2008 1366 International Scholarly and Scientific Research & Innovation 2(5) 2008 ISNI:0000000091950263 Open Science Index, Computer and Information Engineering Vol:2, No:5, 2008 publications.waset.org/11297/pdf