Research On Blind Restoration Technology With Multiple Mode For Target Range Image With Complex Degeneration Condition

Target range experiment is the main way to test the aircraft features and parameters.The images obtained from the optical observation system are the carriers of the parameters mentioned above,so the accuracy of the parameters depend on the quality of the images.Unfortunately,as affecting by some complex factors,the images obtained form the optical observation system are always blurred.So it is hard to measure the parameters from these images.There are two ways to solve this problem:one way is to improve the structure of the optical observation system by using advanced image sensor,which will increase the cost of the system and is not always effective;and the other way is to improve the quality of the images by using digital image restoration technology,which may has a good effect with low cost.So the research on image restoration algorithm for target range experiment images has very important engineering significance.In allusion to the complexity of the degeneration process of the target range experiment images,in this paper,our research on restoration algorithm for target range experiment images focus on three aspects:image restoration algorithms for specific types of blurred degeneration;identification algorithm for both the types of images degeneration process and the model of the aircraft flying status;and the blind evaluation algorithm for restored images quality.Our research can match the type of blurred images with the type of restoration algorithms basically and accomplish refining the quality of restored image automatically.Our research can be summarized as follow:1)As the target range experiment images contain strong noises,which make the images are hard to be restored,the histogram equalization and BM3D denoise algorithm are used to improve the contrast ratio and signal to noise ratio(SNR)of the images,and then we use three type of image restoration algorithm to restore the gaussian blurred image,motion blurred image and out-of-focus blurred image respectively,the experimental results show that these algorithms are effective.2)In order to match the type of blurred images with the type of restoration algorithms,a image blurred type recognition algorithm is proposed,which use aircraft flame and blurred image autocorrelation as features.We use the result of this recognition algorithm to match the type of blur images and the type of image restoration algorithms.3)As the clear target range experiment images can not be obtained,so it is impossible to use non-blind evaluation algorithm to evaluate the restored target range experiment images' quality.In order to solve this problem,a blind evaluation algorithm is used,which combine the noise,residual blur and ringing with the quality of restored images using supervised learning algorithm.The experimental results show that this algorithm is effective.4)As the amount of the images obtained from one target range experiment is very large,it is difficult to restore target range experiment images one by one manually.To solve this problem,we propose an image restoration algorithms with a novel structure,which combine image preprocessing,typical blurred image restoration algorithms,image blurred type recognize algorithm and the restoration image blind evaluation algorithm together effectively.This novel algorithm can restore the target range experiment images effectively and match the image restoration algorithms with the corresponding type of blurred images automatically.By using the output of blind evaluation algorithm of the restored images as feedback,the parameters of the restoration algorithms can be refined and the quality of restored images can be improved.