Deformable templates and image compression

Given two digital images from the same family of images, one of them can be deformed so that it looks similar to the other. This deformed image can then be used to describe the second image. The nature of the deformation and the difference between the deformed image and the second image will help us understand how the two images are similar and how they are different. If the two images are similar enough, then the deformed template will be very close to the second image and storage space can be saved by simply storing the first image and the deformation rather than storing both images.; Algorithms for template deformation are described and their application to the compression of digital images is studied. Two slightly different implementations of template deformation are used which differ only in the choice of basis. Standard methods of image compression are discussed and will be used to compress the error from the deformation and to compress the original images in order to judge the merits of compression using deformable templates. A method for compressing the deformation will be described.; Heuristic arguments will be presented which will help us understand when the deformable template method will yield a deformed template that is close to the original data image. Finally, simulation and experimental results will be presented which show that the deformable template method can improve upon standard methods for image compression, both when the results are evaluated by perceived visual quality and by the mean square error.