Accurate Registration for Low Resolution Images using Wavelet Neural Networks: A Novel Approach

Accurate registration of multiple low resolution images is of central importance in many advanced image processing applications, since capturing of multiple low-resolution images taken of the same scene results in a distortion between each image. Image super-resolution is a typical application where the quality of the super-resolved image is degraded as registration errors increase. In this paper, we have proposed a Wavelet Neural Network (WNN) based image registration, where we are estimating the relative rotation, translation and shift between the observed images and the reference image. We are able to obtain an accurate registration which is very much essential for super resolution image reconstruction. Experimental results shows that the proposed approach is superior when compared to Fourier based registration. Fourier based registration works only for clean images, i.e. images without any degradation, where as our proposed WNN based registration works for severally, degraded images viz. blur and noise.