Study Of Blind Super-Resolution Image Reconstruction Algorithm And Its Application

High resolution images are often required in military remote sensing, public security, digital television, cultural relic protection and recovery, medicine field and so on. Super resolution image reconstruction technique is, in the existing imaging equipment and imaging conditions, using single low resolution frame or a sequence of low resolution images to reconstruct the image with better quality and higher resolution. After super resolution image reconstruction, the obtained image has higher pixel density and contains more details than the low resolution image.This paper focuses on studying 2D super-resolution reconstruction model, blind super resolution image reconstruction and medical image reconstruction based on compressed sensing technique, the main works are as follows:(1) In the traditional 1D reconstruction model,2D image matrix is transformed into 1D vector, making a sharp rise in the amount of storage and calculation in the super resolution image reconstruction process. Under the premise of deformation, fuzzy, sampling operators are kernel-separable, this paper has proposed a super resolution image reconstruction algorithm based on 2D model, which effectively solves the problems of huge storage space and long calculation time required in the classical reconstruction process.(2) Blind super-resolution reconstruction aims at producing a high resolution image without knowing the warping and blur. It has always been the difficult point of super resolution reconstruction. This paper proposes an effective blind super resolution image reconstruction algorithm. Compare with the traditional blind reconstruction algorithms, and confirm that the proposed algorithm achieves well reconstruction in theory and experiment.(3) The medical image reconstruction is an important branch of super resolution image reconstruction, and it is also the hot spot of image reconstruction. Once image’s sparse transform coefficients are available, reconstruct image accurately and rapidly is possible. We apply the proposed SR algorithm to medical image reconstruction, and use the well reconstructed image sequence to display 3D result.