Novel image registration method using gradient based algorithm along with multi-resolution wavelet transform

Image super-resolution is a technique of combining a set of overlapping low-resolution (LR) images to produce a high-resolution (HR) image. This technique has two main steps; the first step is alignment of LR images that should be done with sub-pixel accuracy and the second step is the reconstruction stage. In this paper, we improve the works done beforehand in image registration. The planar motion estimation, using the image gradients, has been modified by replacing a multi resolution wavelet transform instead of Gaussian pyramid for augmenting the registration accuracy.