Motion Blur Image Resroration Theory And Key Technology Based On Total Variation

During motion imaging, since the vibration of the load platform and the complexrocking motion of the camera, there is relative motion between the image and thephotographic medium. It will inevitably lead to the imaging blur, called motion blur, whichgreatly affects the image quality of optical remote sensor and decreases the remote sensingimage resolution and is bad for further applications, such as image feature extraction,automatic target recognition and image analysis. It is strongly necessary to be processed soas to enhance the value of the degraded image.Although there are some hardware solutions to eliminate motion blur, such asimproving the mechanical devices and adding optical devices, it is still limited by thedevice technology, time constraints and alignment construction costs and other factors.Therefore, motion blurred image restoration theory and key technology study has importantapplication value and practical significance.In this paper, with the deep elaboration in current development of image restorationtechnology at home and abroad, we conduct in-depth researches, in the theory of totalvariation, motion blur point spread function estimation and image restoration algorithm.These researches are based on total variation theory of bounded variation space for thebasic framework. The main innovation of the research results are as follows:First, we research the causes of nonlinear diffusion and variational image restorationmodel produces the staircase effect, analysis characteristics of staircase effects on the ofhigher-order variational PDE recovery model and adaptive total variation restoration modelsuppression, establish a coupling gradient adaptive fidelity term total variation imagerestoration model, carry out a feasibility argument theoretically, and discuss the parametersselection criteria. Experimental results show that the proposed model retains the advantagesof adaptive total variation image restoration model to protect the edges, make up the defectthat total variation model does not fully comply with the principles of image processingmorphological, effectively suppress staircase effects in recovery images.Then, we study the motion parameter estimation method on image characteristic in spatial and frequency domain, discuses reasons of arising estimation errors and prove thatthe parameter estimation method based the frequency domain characteristic has higheraccuracy. For the hybrid blur image that information on the frequency domain is aliasing,we proposed a total variation image restoration method that parameter estimation is basedon cepstrum analysis. By analyzing image cepstrum, hybrid fuzzy parameters are estimated.Total variation coupling gradient fidelity term image restoration algorithm is used for imagedeconvolution. Experimental results show that this method can obtain more accurate hybridfuzzy parameter results, effectively inhibited ringing effects caused by motion blurparameter estimation errors.Deeply, we establish a total variation image restoration method by coupling novelimage priors, which are gradient super Laplace global constraints, local image gradientfidelity constraints and blur kernel constraints. An alternating scheme is proposed toestimate the blur kernel and recover image. It uses the standard interior point convexoptimization method to optimize the blur kernel energy function and achieves a blur kernelchange from coarse to fine scale by changing the resolution iterative algorithm. Analternating half-quadratic algorithm is proposed to optimize the non-convex energyfunction for recovery image. Simulation and real blurred image restoration results showthat our method effectively suppresses ringing effects caused by inevitably blur kernelestimation errors, protects image edges, and ease the staircase effect to preserves moreimage detail.Finally, we proposed a single blurred image restoration method based on imagegradient cepstrum analysis. The blur kernel size is determined according to the kernelreproduction information in the image gradient cepstrum, which avoids defects of alternateiterative algorithm that have to pre-set and determine the size of the PSF based on iterationtermination criterion. The phase retrieval algorithm is adopted to estimate the blur kernelfinally that improves the estimation efficiency; then, we use an alternating direction methodfor total variation image restoration to achieve a rapid recovery image deconvolution.Experiments show that the method can overcome slow convergence disadvantages oftraditional alternating iterative algorithms to estimate the blur kernel and recovery image,improve the computing efficiency and gain a more natural visual effects. Study on motion-blurred image restoration theory and key technology based on totalvariation are not only aim to improve image quality, but also provide scientific criterions toimprove performance of the optical imaging system, also can be used for intelligenttransportation, military defense, public safety, etc. many other areas, and provide theoreticaland technical reference for related researches.