Registration Algorithm For Image Super-resolution Technology Research

A clear high resolution image which can provide more details is needed in video surveillance, aerial photography, medical image and so on. It is difficult achieved or high cost that we improve the precision only by raising the hardware devices such as CCD camera . Therefore, there is great significance that we use super-resolution reconstruction methods to improve image resolution.Super-resolution image reconstruction reconstructs a high resolution image from lots of low resolution which have different information by using digital information processing technology. and it overcomes the technical limitations of the hardware system. Currently, image super-resolution reconstruction of image processing technology has become one of the hot area of image processing. As the necessary step of super-resolution reconstruction, image registration effects the greatest impact on the reconstruction, and is the key step , and also is the research focus of this article.In this paper, we describe a several methods of image registration, compare phase correlation method and Keren method specially. Combined with image pre-processing method we eliminate the border scratch, and improve the registration accuracy. We put forward a new image registration method based on the Nyquist criterion and phase correlation method which has good performance on anti-noise. It estimates parameters only based on aliasing-free parts in low resolution image. First, do the Fourier transform for low-resolution images, then carry out the polar coordinate transformation for the magnitude of their Fourier transformation. We compute rotation parameter by construct a function which can turn two-dimensional correlation operation to one-dimensional correlation operation. Then reverse the rotation angle which obtained by rotation parameter estimation, we use phase correlation method for shift parameter estimation only based on aliasing-free parts. We obtain image registration parameters in two steps. Simulation results show that the improvement method has good precise of registration and good effect on image reconstruction.