Research On Motion Blur Video Restored Method

Video is widely used in daily life,public safety,astronautics and other application fields.During the process of acquisition,transmission and storage,the video may be blurred due to various factors.The motion blur due to camera shake and object motion is the most common video blur phenomenon.Blurred video not only affects the visuality,but also loses some information,affecting the extraction and identification of information in the video.Therefore,how to effectively obtain sharp video is important question for long-term attention and research in academia and engineering.Therefore,this research has important academic value and significant engineering significance.Currently,sharp video can be obtained mainly from two aspects that are hardware methods and software methods.The hardware method is to improve the video capture hardware to avoid blur of the captured video.However,this method increases the size,weight and cost of the hardware device.So,this method is less popular.The software method is to use digital image processing technique to recover blurry video and get sharp video.The method has the advantages of low cost and good replacement.In this paper,a lot of in-depth research on the method of motion blur video restoration using image processing technology is carried out.Video restoration is to recover a sharp video from the blurry video.The blurry video loses a lot of important information,so blurry video restoration is apathological blind recovery problem.In order to solve this problem,researchers in the field of motion blur video restoration have proposed many restoration methods.However,because motion blur video not only has spatial information but also time information,the blur in the video is more complicated,which increases the difficulty of restoration.The existing method has certain limitations.Based on the research background,significance and current situation of the domestic and overseas representative motion blur video restoration methods,this paper has carried out the following three aspects on the important and difficult problems of motion blur video restoration methods.(1)A motion blur video restoration method based on inter-frame motion compensation and constraint is proposed.The early traditional video restoration method did not make full use of the spatial and temporal prior information of the video when estimating the blur kernel,and there was a problem that the estimation result was inaccurate and the program running time was long.Considering the full use of video time information and spatial information in the video restoration process,a motion blur video restoration method based on inter-frame motion compensation and constraint is proposed.The method is aimed at restoring a blurry video with constant depth static scene by the camera is moving in a straight line direction,and the blur due to camera shake or fast move.The advantage of this method is that it can restore motion blur video without sharp frames and has a faster program running speed.This method proposes an estimation strategy for blur kernel.The method first estimates the motion vector between the blurry frame and its adjacent frame,and compensates the restored adjacent frame to obtain a motion compensation frame.Then,the motion compensation frame is preprocessed to obtain a derived motion compensation frame,which is input into the improved blur kernel solution model to estimate the blur kernel.The method also proposes a sharp frame solution model.The interframe information constraint of the model uses a time regularization function to constrain the interframe information between the restored result and the motion compensation frame.The model can effectively suppress ringing effect and avoid discontinuous phenomena in the restored video.Finally,the extended fast total variation deconvolution method is used to solve the sharp frame solving model,and the restored sharp video frame is obtained.(2)A motion blur video restoration method with weighted curvelet accumulation is proposed.The blur in the motion blur video that captured under natural state is spatially changed.At this time,there is no special requirement for motion route and captured scene of camera.A new restoration method is proposed for this motion blur video.The principle of the method is as follows.Firstly,because of the randomness of the camera shake,the blur degree of the different video frames in the really blurry video that captured under natural state is different.The same object may be blur in some video frames and sharp in other video frames.Based on this feature of the blurry video,a weighted curvelet accumulation method is proposed to estimate the initial potential sharp frame of the blurry video frame.In addition,when multiple objects are moving in the video that captured by the camera,because the motion state of each object is often different,the ratio of the intra-frame motion vector to the inter-frame motion vector of the different pixels on the different motion states is not same,the ratio we call it as motion vector duty cycle.However,in the existing video restored method,the value is set to the camera duty value or a constant.In fact,when the camera is automatically exposed,the camera duty value is unknown.The spatially varying motion vector duty cycle is set to a constant that causes an error to the intra-frame motion vector in subsequent steps,that final impacts the accuracy of the estimated blur kernel.In order to solve the above problems,a motion vector duty cycle estimation method is proposed.The method can effectively improve the accuracy of the estimated blur kernel.Finally,a new sharp frame solving model is established by the block strategy,the spatiotemporal information and the estimated blur kernel.And finally the restored result is obtained.(3)A motion blur video restoration method based on pixel motion vector duty cycle is proposed.When there are multiple moving objects or large depth changes in the captured motion blur video,a motion blur video restoration method that can estimate the motion vector duty cycle and blur kernel for each pixel of the video frame is proposed.We call the proposed method as a motion blur video restoration method based on pixel vector motion vector duty cycle.The method can restore the blurry video with large different motion vector duty cycles and blur kernels in the adjacent pixel points.Firstly,the motion vector duty cycle and blur kernel of each pixel are estimated by using the maximum and minimum distance clustering method combined with the optical flow and intra-frame projection accumulation model.Then,a sharp frame solving model is established.The model uses the time constraint of adjacent video frames to maintain the temporal continuity of the video.The model introduces a regularization constraint of the motion vector duty cycle to ensure its gradient sparsity.The model introduces the space constraint of the potential sharp frames and optical flows to avoid the ringing effect.The proposed restoration method also constructs a pyramid scheme to achieve fast convergence that can improve the running time of the program.The values of the potential sharp frame,optical flow and pixel-based motion vector duty cycle are update by an alternating iteration method,thereby the sharpness of the restored video is improved.In order to verify the feasibility and practicability of the proposed method,the artificially blurred video and the really blurry video are restored by the proposed motion blur video restoration method and the existing typical video restoration method.Through a large number of experiments,the restoration results are fully compared and evaluated in subjective visual and objective evaluation.The experimental results show that the proposed motion blur video restoration method can effectively restore the motion blur video and obtain better motion blur video restoration results.