Fast Restoration Method For Random Motion-blurred Image Based On Dynamic Optical Transfer Function

Imaging technology has been widely used in military, industrial, civil and manyother fields. However, image will be random motion-blurred because of relative motionbetween the scene and the time-delay imaging system, aero-optical effect or other sea-sons, and imaging performance of the machine vision system will be affected. Imagerestoration algorithm is an effective way to improve quality of image. However, iterationis used, so the shortcomings of the most widely used blind restoration algorithms are largeamount of computation and time-consuming, and it is difficult to meet the fast restorationrequirements in many fields. Linear restoration algorithms have not been practical ap-plication because of the limitations of the existing image quality evaluation function ormethod. Therefore, establishing a new image quality evaluation model and combinedwith the linear restoration algorithm is an effective means of rapidly restoring randommotion-blurred image.The subject of “a fast restoration method for random motion-blurred image basedon DOTF”aims to strictly derive random motion Dynamic Optical Transfer Function(DOTF) model, and use it to accurately describe the degradation law of random motion-blurred image, and then a restoration method which extract DOTF from the blurred imageand combine with the linear filter algorithm is proposed. Iterative operation can be avoid-ed in this restoration method, and the degraded image can be restored rapidly. Theseresearches offer a technology scheme for rapidly restoring random motion-blurred im-age, and it is good for improving real-time imaging performance of time-delay imagingsystem.The main research contents include:(1)In order to solve the problem of the existing image quality evaluation functionor methods can’t be used for rapidly restoring random motion-blurred image, a new eval-uation function was built. Firstly, we rigorously derived Random motion DOTF modelsin the space domain and in the spatial frequency domain separately, and they are mutualverified because the two DOTF models are exactly the same; secondly, Using our DOT-F model, we rigorously derived previously known DOTF expressions for static, uniformmotion and high frequency harmonic vibration, and according to the displacement func-tion ??(??) of the motion image is no limits to any motion type, our DOTF model can bedeveloped for random motion was verified; at last, by strictly derivation, we found thatour DOTF model is the Fourier transform of the existing aero-optical point spread func-tion, it is therefore concluded that our DOTF model can be also applied to describe thedegradation rules of aero-optical random motion-blurred image.(2)The existing blind restoration methods have the characteristics of large compu-tation and long operation time, in order to solve these problems, a fast restoration methodfor random motion-blurred image was proposed. Firstly, the spectrum model of degradedimage was strictly derived, secondly, based on the border added and background equiv-alent static properties of the guide part of the degraded image, a restoration method ofextracting DOTF from the guide part of the degraded image and combining with the lin-ear filter algorithms was proposed; at last, our restoration method was simulated, thecorrelation coefficient between the restored image and original image is 1 proves the ef-fectiveness of our method.(3)The rules of error impacts of our methods effect quality of the restorationimage were analyzed. Restoration models were built for the background is not uniform,the image sensor has dark current, and there is random noise during imaging separately,and the restoration method was simulated for the different models. The simulation resultsshow that: for nonuniform background, quality of the restoration image will be degradedwith increasing of the contrast of the background; for constant dark current, quality ofthe restoration image will be unchanged as long as the dark current subtracted from thedegraded image during the restoration process; for random noise, quality of the restorationimage will be declined with increasing of the noise intensity. These results suggest that inorder to get a high quality restoration image, a background with low contrast and low noiseintensity will be chosen, and the dark current need to be subtracted from the degradedimage during restoration.(4)The fast restoration method and the related theories are verified by experi-ments. Firstly, according to the measurement DOTF curves are agree to the theoreticalones at the same motion parameters, and the changing law of the DOTF curves in thespatial frequency direction with the change of velocity is agree to the theoretical result,the effectiveness of our DOTF model were verified; Secondly, secondly, according tothe correlation coefficient between the blurred image degraded by software and the realdegraded one is 0.9979 and the correlation coefficient between the blurred image with aSynchronous motion background and the one without Synchronous motion background is0.9994, the border added and the background equivalent static properties were verified; atlast, The blurred images were restored by our restoration method and the existing ones, forquality of the restoration image, the GMG and LS of the images restored by our methodare 5.77 times more than those of the images restored by the existing methods, and foroperation time, the cost was cut nearly 90% by our method, it is therefore concluded thatthe advantages of our restoration method are not only improving the restoration quality,but also increasing the operation speed nearly one order of magnitude.