Image Geometrical Analogies

Abstract

Geometric transform (G.T.) of images is a critical operation in commercial television, film producing and advertisement design. All geometric transformation operations are performed by moving pixel values from their original spatial coordinates to new coordinates in the destination image. The traditional algorithms for geometric transformation are time consuming and not accurate. With a very few exceptions, all geometric transformations result in some output pixel locations being missed because no input pixels were transformed there. This paper presents an easy-to- implement and very efficient algorithm for image geometric transform. The idea behind the proposed algorithm comes from the current work on transferring properties between images, where various types of properties (i.e. transformation filters) are to be learned from one source image – and applied to another target image. The results demonstrate the efficiency effectiveness of the proposed algorithm.