Research On Image Motion Deblurring Algorithm

Image deblurring is one of the research focuses in the field of digital image information processing and an important part of image restoration.Restoring an image into an image that is easier to process or has better visual effects has always been the goal of researchers.A common blurring situation is usually motion blur caused by objects moving relative to the camera device.Blurred images can generally be divided into uniform blur and non-uniform blur according to the number of blur kernels contained in the image.Uniform blur refers to the situation in which the global blur of the image is consistent,and a single blur kernel can be used to restore the image better;non-uniform blur refers to an image that contains different blur conditions,and it is difficult to obtain better restoration with a single blur kernel image situation.At present,more research is on de-uniform blurring,and there are many de-uniform blur algorithms with good effects.However,there are only a few studies on non-uniform blur algorithm,and there is a large space for improvement of the existing non-uniform motion blur algorithms in accuracy and speed.This paper mainly studies on the removal of non-uniform motion blur,and the focus of the research is the situation where there are multiple motion blurs in a picture.In this paper,the non-uniform motion blur algorithm guided by soft segmentation is used to improve it.In view of the low accuracy of the soft segmentation de-blur algorithm,the concept of blur degree graph is proposed in this paper,and the blur degree value is introduced into the Mean Shift clustering algorithm.Regions are accurately segmented.For the slow speed of the soft segmentation de-blur algorithm,this paper first predicts the blur kernel in the frequency domain to improve the speed of solving the blur kernel,and then applies it to the adaptive iterative algorithm to improve the accuracy of the blur kernel calculation and reduce the traditional iterative algorithm in deblurring.The number of iterations in the algorithm,thereby improving the speed and effect of the deblurring algorithm.Main work includes:(1)The principle and existing problems of deblurring are analyzed,and the principle and characteristics of commonly used deblurring algorithms are comparatively studied.It is found that the non-uniform motion blur algorithm still has major shortcomings.No matter the algorithm is in terms of image restoration speed or accuracy,there is a large room for improvement.This paper systematically analyzes the research status of de-blurring algorithms at home and abroad,compares the advantages and disadvantages of de-blurring algorithms with better effects at present,and chooses the de-blurring algorithm guided by soft segmentation as the basis for improving the de-blurring algorithm in this paper.(2)Aiming at the problem of low accuracy of the segmentation blur area in the soft segmentation deblurring algorithm,this paper proposes the concept of blur degree map,and takes the blur degree value of a single pixel as a one-dimensional variable of the blurred image,and compares it with the R value of the same pixel.,G and B components together constitute the four-dimensional variable of the blurred image.In this paper,the obtained four-dimensional variables are used as the input variables of the Mean Shift clustering algorithm to accurately segment different blurred regions of the blurred image.The improved Mean Shift algorithm provides an accurate blurred region segmentation result for the deblurring algorithm,which is beneficial to improve the accuracy of blur kernels in different regions,thereby ultimately improving the deblurring effect of the non-uniform motion blur algorithm.(3)Aiming at the problem of slow solution speed in soft segmentation deblurring algorithm,this paper designs a two-step fuzzy kernel algorithm: first predicting the fuzzy kernel in the frequency domain,and quickly and accurately solving the direction and intensity of the motion blur kernel in the frequency domain to improve The speed at which the blur kernel is solved.In view of the problem that the number of iterations of the traditional deblurring algorithm is not suitable for all blurred images,an adaptive iterative refinement deblurring algorithm is designed in this paper,and the number of iterations of the deblurring algorithm is optimized based on the fuzzy kernel of frequency domain prediction and the degree of image blurring.,in order to reduce the redundant calculation of the algorithm,improve the speed of the algorithm and the deblurring effect.(4)Compare and analyze the effect of different algorithms to remove non-uniform motion of images.This chapter compares the processing results of the improved algorithm and similar algorithms horizontally.The comparison results show that the improved fuzzy region segmentation algorithm in this paper can improve the accuracy of fuzzy region segmentation,the designed fuzzy kernel solving algorithm can accurately solve the distribution of fuzzy kernels,and the overall deblurring algorithm in this paper can improve the speed and accuracy of the deblurring results..The peak signal-to-noise ratio PSNR of the restored image has been significantly improved,and the processing speed of the algorithm has been significantly improved,which shows the correct restoration effect and processing speed of the algorithm in this paper in the correct restoration and processing speed of blurred images.