Research On The Practical Technique Of Image Blind Restoration

The research of image blind restoration not only is important theoretically, but also has urgent demand in actual applications, and it has become an active topic in recent years. However the research on this subject is still elementary, and systemic analytical methods and filter designing methods have not come into being yet because of the shortage of available information, foundational theories and immature applied researches. Thus it is of great importance to research further. On the other hand, the existing techniques do not have strong applicability or satisfactory effect, and most methods only have theoretical significance, so they cannot be used practically. Valid paths to resolve the actual problem have not been formed yet.This dissertation proposes a practical frame for image blind restoration with purpose to improve the practicability of image blind restoration. After getting observation images, the frame can automatically identify the blur types and adopt corresponding methods of blind restoration, experiments showed satisfactory results.This dissertation is carried out in the following aspects.1. Based on a systemic research of image blind restoration, it proposes a practical processing frame for image blind restoration.2. It realized blur-type recognition utilizing the cepstrum method. It identified blur types as motion blur, outfocus blur and other blur and devised a method to identify blur parameters of motion blur and outfocus blur, then to restore the blurred image.3. It carries out an iterative restoration method, which can process other blur types that do not belong to motion blur and outfocus blur and raises the accuracy of the estimated results of PSF support size by using MCSC criterion.4. With regard to the ringing effect that frequently appears in image blind restoration, it proposes a postprocessing method, which can eliminate ringing effect effectively and preserve much of the image details. The post-processing method improves the quality of restoration image remarkably.5. In response to the characteristics of blind restoration, it adopts an image quality assessment scheme to separately assess the ringing effect and definition of the image being restored, so as to avoid the interference normally introduced into a single parameter image quality assessment system by ringing effect.