Coded Aperture Techniques For Catadioptric Omni-directional Image Defocus Deblurring

With the rapid development of the computer science and the informationtechnology, great progress has been made in the field of the digital image processing. Atthe same time, it also brings much convinence to the human being’s life. Nowadays,since the “computational photography” technology develops dramaticly, the field ofdigital image processing steps into a brand new era.The catadioptric omni-directional imaging system takes advatanges in the wideview fiels over traditional camera and widely used in many fields. However, Thedefocus blur in catadioptric omni-directional imaging is caused by large apertures andmirror curvatures. More recently, this problem becomes obvious when introducing highresolution sensors.In order to overcome this drawback, this work proposes a simple modification to aconventional catadioptric system that allows for the recovery of an all-focusomni-directional image. The modification is designed by inserting a patterned occluderwithin the aperture of the camera lens, creating a coded aperture. Next we analysis theomni-directional imaging system, introduce a criterion for blur discriminability whichaims to design the preferred aperture pattern. In the end, by using specificdeconvolution method, we can recover an all-focus omni-directional image fromphotograph(s) taken by the camera with coded aperture. The resulting sharp image canbe combined for various catadioptric applications, including omni-directionalmonitoring systems, intelligent omni-directional systems and robotics, etc.This thesis focus on blur of catadioptric omni-directional imaging system and themain works include:The statistical model of the defocus blur in catadioptric omni-directional imagingsystem based on coded aperture techniques. The blur of the catadioptricomni-directional image has a relationship with the design of the imaging system. Byanalyzing the parameters in the catadiopric imaging system, taking the coded aperturetechniques into concideration, we can get a statistical model of the system.The PSF calibration of the catadioptric omni-directional imaging system. Toidentify the blur scale of the blurry image, we need a PSF calibration of the catadioptricomni-directional image formation system. The calibration plays an important role in therecovery of the blurred image. This thesis brings out a new calibration for theomni-directional imaging system and makes simulation with Zemax.The image deblur algorithm. The deblur algorithm is a deconvolution process inessence. However, the deconvulution process is an ill-problem. By introducing andcomparing the two famous Richardson-Lucy deconvolutoin and Wiener deconvoluitoin,this thesis put forward a new deconvolution based on sparse image prior, and shows some experiment results. At last, the entropy and the average gradient indicate the goodquality of the recovered images.