Blind Image Restoration Of Motion-blurred Images

Motion blur can be caused in many occasions in the process of scientific work, such as the photo shot at the plane or space craft, the photo of object while camera dithering. The relative motion between the image recording device and the photographed object in the short time of exposure is called motion blur. It brings great difficulty to the follow-up image analysis and target recognition. So the restoration of motion blurred image becomes necessary. Most of restoration approaches require the prior knowledge of blur and some features of noise. However, such an assumption may not be satisfied in practice. Recovering an original image from several degraded versions without such knowledge is called as blind image restoration.In this paper we fully research on the restoration of motion blur caused by uniform rectilinear motion. Through analyzing the imaging principle of motion blurred image, the degraded model and restoration model are built. What's more, the key to restore motion blurred images is to obtain the parameters of point spread function (PSF). While the existence of noises affect the restoration results seriously. Our research work mainly includes three aspects: estimating the degraded model and identifying the parameters of PSF, choosing the appropriate restoration algorithm, finding a more significant method to decrease noise.In order to identify the parameters of the PSF accurately, an identification method based on cepstrum domain is presented in this paper. The gray-scale transformation and edge detection of the cepstrum is employed to extract blur length. The Radon transformation of the cepstrum is used to identify the blur direction. The experimental results covering 130 artificially blurred images and real images show that the proposed method is accurate and of robustness. Final restoration is achieved by using classical Weiner filter, the restored images also validate the usability and accuracy of the proposed method.Considering the motion blurred image with noises, a joint motion burred images with noise restoration algorithm based on Weiner filter and wavelet transform is presented. This algorithm was on the basis of the identified PSF parameters. The Weiner filter was employed to proceed with the primary filter. The experimental results show that the proposed method is more efficient to the noise than Weiner filter, the PSNR of the restored image is increased by as much as 2.2378dB. An improved semi-soft threshold method of the wavelet transform was used to filter the noise. It changes the linear relationship among wavelet coefficients into exponential form. So the errors of reconstructing signals can be reduced.