This thesis describes an investigation into the uses of small computers in digital image processing with specific reference to the two-dimensional Fourier transform. The motivation for this work stems from a natural curiosity in determining the effectiveness and limitations of minicomputers in an area where large storage and high speed requirements predominate, and more importantly in developing effective image transformations suitable for small machines with the use of future generations of high speed microprocessors in mind. Chapters one and two are introductory in nature. Chapter one serves as an introduction to the complete range of image processing activities. It is also hoped that this chapter provides a concise summary of image processing techniques of interest to the user of small computers, which is lacking elsewhere. The necessary mathematical tools required for an understanding of digital image processing using orthogonal transformations are developed in chapter two, partly in an historical context of analogue Courier processing. The remaining chapters describe the development of an image processing system, using a typical minicomputer, based on the use of the two-dimensional fast Fourier transform and on the application of this system to the processing of side scan sonar imagery. Of particular interest is the problem of measuring imaging errors resulting from processing, or of measuring observable differences between images; an approach to this problem utilising some knowledge of the way in which a human observer's visual system composes images, and using the two-dimensional Fourier transform, is described in chapter four. There is much scope for further research in this topic. The importance of repeatable reproduction of digitally processed images is frequently referred to, and consequently the practical apparatus and photographic methods used are also briefly described in appendices.
|Date of Award||1977|