Research On The Algorithm Of Super-resolution Reconstruction

Super-resolution reconstruction is to reconstruct a high resolution image beyond its cut-off frequency with more details and message. The traditional image reconstruction methods use a low resolution image. Super-resolution reconstruction can reconstruct a frame of image with high resolution by using low resolution images which have strong relationship between each other.So far, the technology of Super-resolution image reconstruction has been extensively applied to remote sensing, military, public security, computer vision, medical imaging, multi-media electronic consuming, image compression and so on. Therefore, this technology has great prospect with significant application value.This thesis firstly introduces the developmental process of Super-resolution image fusion as well as all kinds of algorithms basing on both frequency domain and spatial domain. Two of the most successful algorithms will be studied in depth basing on the full analyses and comparisons of all these algorithms.Two key steps of this technology are the establishment of observation model and image registration (motion estimation). Thus, this thesis secondly sets up the observation model, and then provides a detailed introduction to the principle, classification, and manipulation of image registration. This thesis carries out experiments of two registration methods which are based on the image gray and Fourier transformation on the MATLAB6.1 platform and compares the both methods.Thirdly, the categories of interpolation are analyzed. Three usual kinds of methods are studied and the affection of methods is showed by experiments on the MATLAB6.1 platform. The bicubic interpolation is superior to the nearest neighbor interpolation and the bilinear interpolation.At last, this thesis focuses on the study of the principle, realization and steps of the Iterative Back Projection method. Meanwhile, this thesis analyzes the factors impacting on the reconstruction results(Iterative numbers and the frames of low-resolution images)through experiments on the MATLAB6.1 platform and compares the interpolation method with the Iterative reconstruction.