Study On Theory And Algorithms Of Optical Synthetic Aperture Image Restoration

The wide applications of earth observation missions bring an urgent need for imagingtelescope with a higher spatial resolution. However, the light diffraction effect restricts theresolution of optical imaging system, while the angular resolution of conventional opticalimaging system is limited by the wavelength and aperture. The only way to improve the angularresolution of system is to enlarge its aperture size. On the other hand, restricted by the factors ofmanufacture techniques, load volume and weight, the launch diameter of space-based remotesensing imaging system is supposed to be below4.5m while the design diameter of ground-basedastronomic telescope is limited within about10m. Optical synthetic aperture imaging system,which is composed of multiple sub-apertures, is a novel type of sensor that can break through theaperture limit and realize the ultra-high resolution imaging. A number of sub-apertures arearranged in a certain way to meet the co-phase conditions and realize the interferometric imagingon the focal plane, and by virtue of these, this imaging technology is mange to achieve theresolution of equivalent single aperture telescope.Compared with equivalent single aperture system, the partially filled collecting area ofsynthetic aperture system reduces the response of low and middle spatial frequencies, whichinevitably lead to the decline of image contrast and bring blur and degradation. Consequently,the post-processing such as image restoration and reconstruction has to be implemented to obtainhigh-resolution remote sensing images and take full advantage of synthetic aperture imagingsystem. To meet this aim, this thesis carries out the study on synthetic aperture image restorationtheory and algorithm, prepare the technical supports for the development of next generationnovel sensor and provide fundamental theories and algorithms for corresponding data processingsoftware system.The theory and algorithms of synthetic aperture image restoration is the main contents ofthis thesis, the main achievements are as follows:1The study on the imaging principles of optical synthetic aperture system has been carriedout. The characteristics of point spread function and modulation transfer function have beenanalyzed, the influences of beam combining errors on the imaging system and images have beendiscussed and the thorough research and analysis have been put on the imaging model, spectralresponse and noise type of synthetic aperture system.2The experiments and simulations of optical synthetic aperture imaging have beencompleted, and the corresponding simulation and post-process software has been developed. Theimages with various array structures and fill factors have been acquired on the optical syntheticaperture system testbed. The pupil function, ideal point spread function and modulation transferfunction of different arrays have been simulated and compared with that of experimental images.The piston error, which is the most important one among beam combining errors, has beensimulated and analyzed; the simulated images with the max piston error have also been obtained.3The improved direct blind deconvolution algorithm to reconstruct the optical syntheticaperture image has been analyzed and studied. Pointed to the characteristics of synthetic apertureimaging system, the modified algorithm calculates the ideal point spread function, carries outFourier transform, weighted-sum the spectrums in representative directions to replace the original optical transfer function estimation. The experiments to the images with piston errorprove that the proposed algorithm performs better than Wiener filter and original algorithm onthe reconstruction accuracy and restoration details and the validity of the algorithm is alsoverified.4The maximum likelihood method has been modified and realized. According to the noiseanalysis for synthetic aperture imaging, the maximum likelihood method based on mixedGaussian-Poisson noises is more accordance with the practical imaging environment. However,the algorithm has some defects and lacks of stabilities. The methods to adaptively estimate theparameters of noises and point spread function have been respectively proposed to improve theoriginal method. The images restored by modified method own a significant improvement onsignal-to-noise ratio and average contrast, and the simulated image results are similar to that ofWiener filter algorithm.