Research Of Motion Blur Imaging Model And Super-resolution Reconstruction

Usually we regard image as a two-dimension distribution, and imaging system is a linear system as well. The better we get a good knowledge of image and imaging system, the better to research reconstruction algorithm. The system of image and imaging is briefly presented in the front of this thesis, followed by the principle of CCD imaging system. Esepcially, we analyze the function of CCD in motion imaging system of tiny aerocraft, besides, we introduce the standard of quality evaluation.The degradation is space variation when we obtain a number of images damaged by blur and noise in high speed motion environment. We need to establish imaging model before the high quality super-resolution reconstruction. Therefore, we present imaging model, open out the essence of motion blur, and propose the design framework of super-resolution reconstruction, and point out the key method.By the way of research of imaging system and hierarchical search principle, we have knowed that coarse resolution can transfer the matching location to the fine resolution location. If we give more iterative numbers on coarse resolution than the fine resolution location, we can quicken the speed of algorithm. Therefore, we propose a couple of different registration algorithm which based on six-parameter and four-parameter affine transformation, and consider many frames which contain different rotary angle and motion blur degree. Both of the algorithms are precise, and the registration speed of six-parameter algorithm improved by 50 percents relative to traditional algorithm, the four-parameter algorithm improved by 75 percents when rotary angle less than five degree, but as much as six-parameter algorithm when rotary angle increase. The simulated experiment results are provided to illustrate the performance of the proposed algorithm is rapid convergent,precise and robust. It is significant for us to solve super-resolution reconstruction in real-time way.The super-resolution reconstruction will become the main direction of image restoration, and the POCS method is the framework of iterative algorithm. We present the POCS super-resolution reconstruction algorithm of motion blur image in detail, which contain imaging model, interpolation of reference frame and kernel restoration algorithm. Follow by the analysis of ringing effect which always happen at the edge of image. In order to weaken the direction of edge vary, we propose that adopt smooth filter on grey image, which improve the quality of image obviously. As we all known, the reconstruction image exist blind area when motion bur exceed the half size of original image. Thus, we analyse the condition of the length of motion bur exceed the half size of original image and the PSF is not square matrix, then we open out the limitation of POCS algorithm, and propose an extended large blur reconstruction algorithm of POCS which based on the different processing method of low resolution reference frame. then simulated experiment results are provided to illustrate the performance of the proposed POCS algorithm is widely applicable, reliable and robust.Image model is important for image restoration, if we research the essence of image model carefully, we can obtain more impactful method of image restoration. Then we research the Markov Random Field model and some technological framework of image restoration primary, analyse the restoration problem based on maximum a posterior probability theoretically, and present the ICD optimization algorithm at last.