Research And Implementation On Motion Deblurring From A Single Image

The restoration of blurred images is an important branch of digital image processing.As an important information source,image plays an important role in traffic monitoring,criminal forensics and meteorological monitoring.It is a very common phenomenon that the motion blur caused by the relative displacement of the imaging target and the imaging device is a phenomenon that occurs when the image information is degraded due to various adverse factors.As the motion of fuzzy degeneration of the scene is often not repeatable,therefore,from the single motion blurred images to restore the potential information in the theory and practical application is of great significance.The restoration of motion blurred images mainly has two aspects: one is the estimation of blur parameters,and the other is the known blur parameters after deconvolution recovery.The causes of motion blurred images obtained in practical applications are more complex,and in addition to the factors of motion blur,they usually contain random noise.Noise has a great impact on image restoration.In this case,the image is often subjected to de-noising before restoration.However,no matter what kind of denoising algorithm,at the same time,it can destroy the high-frequency information such as the edge of the image,and a new method of motion blurred image restoration based on wavelet transform is proposed.The main contents are as follows:Firstly,the research status of motion blurred image restoration technology at home and abroad is summarized.Then some basic theoretical knowledge of motion blurred image restoration are introduced.The model of motion blurred image degradation and the composition of point spread function are described.Several typical recovery algorithms and several commonly used quality evaluation parameters are listed.The first key step in image restoration is to estimate the blur parameters to be estimated.The accuracy of the parameter estimation directly affects the quality of the final restored image.Based on the study of motion blurred image spectrum,the characteristics of its spectral image are found,and the secondary spectrum is obtained.The direction of motion blur can be determined more accurately by using the secondary spectrum.The image is then rotated to the level,using the autocorrelation curve of the differential image to determine the scale of motion blur.The fuzzy image in practice usually contains noise.By analyzing the influence of noise on image restoration,a motion blurred image restoration method based on wavelet transform is proposed to solve the contradiction between the noise removal and the edge of the preserved image.By introducing the wavelet transform,the high and low frequency information of the image is separated,and the high frequency coefficients of the wavelet decomposition are denoised by the improved non-local mean filtering algorithm.The low-frequency coefficients of the wavelet decomposition are used to estimate the point spread function,and the wavelet coefficients are reconstructed by wavelet transform to obtain the reconstructed image.This method balances the contradiction between noise removal and edge preservation to a certain extent.