Generalized Scale

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 and diagnosis of various §human diseases. 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 several image §processing tasks.