Automatic Image Registration Using FFT Based Technique

Abstract

In field of computer vision, automatic image registration between two images at different viewpoints is a key issue. We frequently have two or more images of the same scene obtained at different times by the same sensor or by different sensors with different viewpoints at the same time. These images are often designed as scaled, rotated, and translated replicas of one another.These picture pairs must be recorded for a variety of re- alistic and substantive interpretations. Multi-spectral and hyperspectral data registration in satellite images, computed tomography (CT), and magnetic resonance imaging (MRI) image registration are examples of medical imaging applications. This thesis presents efficient image matching technique with translation, subpixel translation. Fast fourier transform technique is most powerful area based technique which involve translation, rotation and other operation in frequency domain. In this thesis, we discussed different methods of subpixel estimation. In addition, we test our method’s efficiency under a variety of noise conditions. Our main aim is to devise a method for estimating translation and rotation using FFT techniques. PYTHON programming is often used in our implementation