Motion Blur Image Restoration Technology Research

Image restoration is one of the important topics of digital image processing, but also adifficult issue, which aims to improve the quality of the given image, the original image isreconstructed or restored through some a priori knowledge of the process of degradation. Therestoration of motion blurred image is a kind of more common and more difficult in imagerestoration area, which is widely used in the fields of aviation, military, road monitoring,criminal investigation and medical as well as other fields, the research has important practicalsignificance. Many scholars have carried on in-depth study, and achieved significant results, avariety of algorithms for the motion blurred image restoration have been proposed, but thereare still some problems for the study of this subject, which have not been satisfactorilyresolved yet. The motion blur caused by relative motion due to camera equipment and objectsat the moment of exposure, the process of recovery is firstly to establish the model of imagedegradation, and then the restored image can be obtained by calculating the inverse process ofimage degradation.This thesis firstly introduces the background of research and significance of imagerestoration, the current situation of development and application. Secondly, the basic theoryof image restoration is researched, analysis the causes of noises generating in the blurredimages, as well as classification and characteristics of noise are discussed, and elaborate amodel of fuzzy image degradation, especially the model of uniform linear motion blurdegradation, as well as introduces the principles of image restoration. Several classicalalgorithms of fuzzy image restoration are analyzed detailedly, and then the advantages anddisadvantages of each algorithm are summed up. An improved method of estimating the Kvalue automatically is researched for the estimation of the K value in the Wiener filter, so asto avoid effectively the problem of the range of K value is unsuitable, and K value can beestimated accurately. Again, the relationship between the uniform linear motion blur and thegeneral motion blur is discussed, the proposed algorithm—an improved algorithm of uniformlinear motion blurred image restoration is put forward based on summarizing the research offormer scholars, and experimental simulations are carried out by using MATLAB, the idealresults is obtained. Finally, the criteria for evaluating the quality of image restoration areresearched, and then the results of recovery with the method of this dissertation are evaluatedsubjectively and objectively, besides compared with the results of ordinary Wiener filter, theWiener filter of improved K value estimated automatically, the results shows that theproposed algorithm obtains more higher evaluation in both subjective and objective aspect. Overall, analysis by observing the recovery results and experimental data to concludethat the proposed algorithm is superior to the Wiener filter of improved K value estimatedautomatically, more than the ordinary Wiener filter, can achieve better recovery results, is amore effective one for uniform linear motion blurred image restoration, but there are somelimitations, for example, it does not apply to the occasions which requires high real-timeprocessing, more deeper research is needed.