This thesis is concerned with image reconstruction problems encountered in signal or image processing and with the numerical methods used to tackle these problems. We present a theoretical basis for image reconstruction methods and discuss their implementation within the framework of an underlying linearity assumption. The linearity assumption reduces the problem, in most practical situations, to solving a system of linear equations. This simplifies the investigation of the difficulties inherent in the problem. However, most existing methods for mitigating those difficulties have a common estimation structure, called "regularization" in spite of their apparent variety. This is used here as a unifying framework for understanding this rather complex field of image processing and in this setting, image reconstruction methods are investigated. The major emphasis is given to consideration of methods which are based on the principle of the maximum entropy, known to be a useful tool in image processing. We attempt to improve the existing maximum entropy methods used in the literature and propose two new approaches for generating maximum entropy images. The performances of all methods introduced in the thesis are tested for one and two-dimensional images with a variety of noise levels and are compared to each other. In addition, applications to real observations in x-ray astronomy and in nuclear medicine are presented
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:543066 |
| Date | January 1991 |
| Creators | Ustundag, Dursun |
| Publisher | University of Birmingham |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
Page generated in 0.0024 seconds