Study Of Infrared Motion Blurred Image Restoration Based On The Low Rank Matrix Recovery

Image is the main way to obtain information for human, and it occupies an important status in modern society. Infrared technology is used widely because of its advantages, such as far distance, good anti-jamming, strongly penetrate dust haze, and the capability of working at all time and under all-weather conditions. Compared with the visible image, infrared image usually has some disadvantages, such as low SNR, image blurring and low contrast. Therefore, it has a very important significance to recovery infrared image to satisfy the visual needs and the engineering application. It is inevitable that image blurring due to camera equipment jitter or target fast-moving in the process of obtaining infrared image, which includes the imaging process of observing moving target on the stationary platform, such as the process of electronic eye shooting the illegal traveling vehicles, the process of ground surveillance equipment monitoring the air target, and it also includes the imaging process of observing moving target on the motion platform, such as remote sensing imaging process on satellite or airborne platforms, the process of rotating mechanism capturing and tracking dynamic target, etc. The recovery of infrared motion blurred image has become an important research subject in many fields, such as astronomy, meteorology, military detection, traffic monitoring, medical, etc. Therefore, it has important value to study infrared motion blurred image restoration problem.This thesis mainly studies the blind restoration problem for infrared motion blurred image. Firstly, this thesis analyses the purpose and significance of this research subject and introduces the research status in the field. Secondly, the thesis researches the degradation model of the infrared degradation image and several classic image restoration algorithms, and then introduces some image quality assessment methods in detail. Thirdly, the thesis studies several common estimation methods of fuzzy parameter, such as cepstrum, Hough transform, differential autocorrelation and the direction differential method. Fourthly, the thesis researches an image blind restoration method which is based on gradient domain, the algorithm can recovery nonlinear motion blurred image. Finally, the thesis puts forward a blind restoration algorithm of infrared motion blurred image, which is based on low rank matrix recovery, and introduces local polynomial approximate confidence interval cross method, then make it as the input matrix of low rank matrix recovery method, and averaging the low rank parts of the output matrix, thus we can get the final infrared restored image.It is simulated through computer software platform. The simulation results show that the algorithm in this paper is more effective than traditional restoration algorithms.