ALGORITHM FOR RESOLUTION IMPROVEMENT OF IMAGES

Abstract

The process of constructing high resolution image from low resolution images is called superresolution. To achieve these algorithms precise registration algorithms to align low resolution images are needed. The low resolution images are Undersampled which have the problem of aliasing and misregistered. So as to overcome the problem of aliasing it uses the low frequency aliasing free part of the image is used for registration algorithm. The registration algorithms that are used are broadly divided in to spatial and frequency domain. Here new type of frequency domain method is used. This registration method uses a one dimensional correlation where as other methods uses two dimensional correlation. So obviously the execution time of the algorithm will be lesser compared to other methods. A high-resolution image is then reconstructed using cubic interpolation. After implementing the output image of this method is compared with the original high resolution image, and also with the conventional interpolation techniques and errors are compared.