The Super-Resolution Restoration Of Motion Blurred Images Based On Radon Transform

As an important branch of the digital image processing, Image restoration techniques have a wide range of applications in these fields of production and living, transportation, aerospace science and technology and so on, and have gradually become the hotpot of modern computer technology. By investigating the current restoration of motion blurred images, the inaccurate estimates of point spread function and inappropriate choices of image restoration algorithms would directly affect the ultimate restoration. Therefore, an algorithm combined Radon transform with super-solution image restoration is proposed to research on the restoration of motion blurred images.The problems on the restoration of motion blurred images are mainly researched in this thesis, and study is deeply made for the estimation of point spread function of degraded images and the restoration algorithm.Firstly some basic theory for the restoration of motion blurred images are introduced, and then several common methods are detailedly analyzed to estimate point spread function of motion blurred images, also the brief introduction is presented to some of the traditional methods for image restoration, at last the objective and subjective evaluation criteria for image restoration quality are proposed.The inaccurate estimation of point spread function and the unfavorable restoration exist in the existing literature. To solve the problems, an algorithm for super-resolution restoration of motion blurred images based on Radon transform is proposed by combining Radon transform and MPMAP algorithm (RMPMAP). Compared with other estimation methods of point spread function, a detailed analysis and study on the degradation process of motion blurred images are made through Radon transform, the formula to calculate motion blurred direction and motion blurred length is derived, and good accuracy which is suitable for any size of image is also obtainedn in the thesis. Then, the Poisson-MAP super-resolution image restoration algorithm with Markov constraint to restore the degraded images is applied, and can effectively eliminate or reduce the noise and oscillation effect comparing with traditional restoration filtering method. The experimental simulations show that the RMPMAP algorithm achieves better results, and the PSNR can be increased by about2dB.In order to obtain better quality of restored image, the MPMAP algorithm is improved. As a multi-scale and multi-resolution algorithm with highly anisotropic, the second generation Curvelet transform has strong abilities to express the edge information in images, and a wide range of applications in the field of image denoising. Therefore, the improved super-resolution restored image algorithm is proposed by combining MPMAP algorithm with Curvelet, that is CMPMAP. The algorithm combines advantages with both, and greatly enhances the abilities of suppressing noise and oscillation. Many simulation experiments proves that, compared to chapter three algorithm (RMPMAP), the PSNR can be increased by0.2to1.2dB, especially the more noisy, the more rejection capability, and the PSNR be improved more obvious.