Abstract
Conventional Digital Image Processing reaches its limits in those cases, where spectral characteristics (grey levels, colours, spectral bands) cannot be properly associated with specific object — or structural properties. Examples are manifold; Typical scanners in medicine — like ultrasound, CT, MRI, or even light microscopy — frequently produce images, where the grey level distribution is not related to the global or local image characteristics. In other words, simple thresholding algorithms are in such situations not powerful enough to reveal structural properties.
The clue for solving the problem is to take the contextual information into account. This is exactly the key feature behind the GOP theory developed by Prof. G. Granlund at the Technical University in Linko**, Sweden: The information contents of every pixel is not only dependent of its own spectral value, but related to its embedding in neighbourhoods of varying size. The processing is hierarchical, with different levels being combined either top-down, as well as bottom-up. The resulting pixel values are represented as 16 bit in a polar coordinate system, with the low byte giving the contextual feature (e.g. orientation), and the high byte the certainty (“strength”) of that feature Specific hardware and software has been developed in order to realize the ideas of Prof. Granlund.The result, the GOP 300, may be trained, using spectral and contextual features, in order to identify and classify virtually any type of texture or structure that characterize the objects or regions of interest in a tissue.
Besides classification, applications include 3D reconstruction, context — controlled image enhancement (an operation analogue relaxation), and postprocessing within the field of Magnetic Resonance Imaging, Computer Tomography, PET investigations, Nuclear Medicine, Ultra sound examinations. Mammography, as well as light and electron microscopical applications
In the following, the major features of the GOP in Magnetic Resonance Imaging are presented.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Similar content being viewed by others
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 1987 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Schwarz, H. (1987). Contextual Image Processing in MRI — Applications. In: Meyer-Ebrecht, D. (eds) ASST ’87 6. Aachener Symposium für Signaltheorie. Informatik-Fachberichte, vol 153. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-73015-3_47
Download citation
DOI: https://doi.org/10.1007/978-3-642-73015-3_47
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-18401-0
Online ISBN: 978-3-642-73015-3
eBook Packages: Springer Book Archive