Generalized scale: Theory, algorithms, and applications in image analysis
Magnetic Resonance (MR) Imaging (MRI), a non-invasive method for imaging the human body, has revolutionized medical imaging. MR image processing, particularly segmentation, and analysis are used extensively in medical and clinical research for advancing our understanding of the various diseases of the human body, for their diagnosis, and for developing strategies to treat them. These efforts face two major difficulties—the first due to image intensity inhomogeneity present as a background variation component, and the second due to the non-standardness of the MR image intensities. ^ Scale is a fundamental concept useful in almost all image processing and analysis tasks. Broadly speaking, scale related work can be divided into multi-scale representations (global models) and local scale models. In this thesis, we present a new morphometric scale model that we refer to as generalized scale which combines the properties of local scale models with the global spirit of multi-scale representations. We contend that this semi-locally adaptive nature of generalized scale confers it certain distinct advantages over other scale formulations, making it readily applicable to solving a range of general image processing tasks. In the context of this thesis, however, we limit ourselves to addressing two issues that are relevant to MR image processing, namely, (1) correcting for image intensity inhomogeneity (referred to as inhomogeneity correction), and (2) correcting for the non-standardness of the MR image intensities (referred to as intensity standardization). ^ While the importance of intensity standardization and inhomogeneity correction in MR image analysis is well established, their behavior has been studied only in isolation, and their influence on each other has thus far not been addressed. Part of this thesis work is centered on studying the possible effects of one method on the other, in order to find the sequence of inhomogeneity correction and intensity standardization operations that will produce the best overall image quality. ^ The results of extensive quantitative and qualitative evaluation of our methods on nearly 6000 3D clinical and phantom MR data sets are also presented. ^
"Generalized scale: Theory, algorithms, and applications in image analysis"
(January 1, 2004).
Dissertations available from ProQuest.