International Journal of Computer Applications (0975 – 8887) Volume 72– No.17, June 2013 21 Novel Approach to Estimate Motion Blur Kernel Parameters and Comparative Study of Restoration Techniques Kishore R. Bhagat Dept. Digital communication NRI Institute of Information Sci.&Technology,Bhopal, India Puran Gour Dept. Digital communication NRI Institute of Information Sci.&Technology,Bhopal, India ABSTRACT Motion blur occurs due to the fact,that during exposure time there is movement of object or camera or both. Removing of blur is always challenging for image processing field because one has to estimate the motion blur which can spatially differ over image. Motion blur is simply an undesired effect. Restoration of blur image is very important in many of the cases like identification of criminal face in blurred image.In restoration of motion blur the knowledge of the point spread function (PSF) plays a vital role. This paper present a novel approach towards estimation of parameters like motion blur angle and the motion blur length which defines the PSF. Both of these parameters are used to restore the blurred image. Furthermore paper discusses the comparative study of different restoration techniques. The experimental result shows estimated blur angle and blur length are very close to theoretical value and the blur images with natural and artificial noise are successfully restored. General Terms Blur kernel parameters, Image fusion techniques, Keywords Image restoration, Image fusion PSF,Spectrum, Wiener. 1. INTRODUCTION Motion blur eventually causes degradation of image quality. The motion blur can be removed by minimizing the shutter exposure time but in case of low light situation this will causes unavoidable tradeoff of increased noise. One of the frequent reason of motion blur is camera shake. We can address this problem by using some mechanical means like camera can place on tripod stand. Secondly the object movement in the scene causes the motion blur. This type of blur is harder to control so it is often desirable to remove it by post synthesis of blur. Images may be blur due to improper focusing of lens, atmospheric turbulence, undesirable working of optical systems, relative motion between the camera and scene. Hence in the restoration of noisy and blurred images knowledge of blurring system is important. Motion blur is defined with two essential parameters called motion blur angle and motion blur length. An Interactive Deblurring Technique for Motion Blur in which Segment based semi- automated restoration method is proposed using an error gradient descent iterative algorithm[1].R. Lokhande, K.V. Aarya and P.Gupta [2] successfully proposed the algorithm to determine the motion blur PSF parameters.Numerous methods for estimation of blur parameters have been proposed in literature[3,4,5,6,7]. Fabian and Malah[4] discussed about the improvement in sensitivity of Cannon's method. Huiji& Chaoqiang Liu[5] introduce a hybrid Fourier-Radon transform to estimate the parameters of the blurring kernel with improved robustness to noise over available techniques. Gennery has discussed the parameters of blur function (PSF) in spectral domain[7]. Taeg Sang Cho has discussed about Blur Kernel Estimation using the Radon Transform[8]. Alex Rav-Acha, Shmuel Peleg[10] describes how different images, each degraded by a motion blur in a different direction, can be used to generate a restored image. Y. S. Chen and I. S. Choa describes transformation of the enhanced spectral magnitude function in cepstral domain[11] and the blurred image is pre- processed to remove the noise using Spectral Subtraction method[12] , whereas Cannon[17] has identified the point spread function (PSF) parameters in the power cepstrum of the image by inspecting the negative peak. In most of the blur kernel parameter identification methods the blur motion is consider along the horizontal direction, in practice which is not always the case.The method of detecting the blur angle by rotating the coordinate system following by computing its 1-D spectrum and inspecting the peaks and valleys in it is discussed by Li and Yoshida[9]. This paper suggest the approach to find out motion blur PSF parameters, i.e. blur angle and blur length, in frequency domain. After getting the approximated parameters those are used in restoration of image. The different types of filters are used for restoration of images and have been compared in this paper. Further the wavelet based fusion is used for betterment of results. Further this paper is organized as section 2 contains the Image Acquisition Model. Algorithms for identification of motion blur parameters have been explained in Section 3. The comparison of restoration method is given in Section 4. Section 5 contains the wavelet based image fusion. The experimental results are given in section 6. Finally conclusion is given. Following figure shows the propose stages to remove the motion blur.