Post-processing of segmented volumetric image datasets

Abstract

This work concerns the problem of post-processing segmented volumetric images. Such images can arise in fields as diverse as medical imaging, material science, geology and other fields. An application in which to experiment with different post-processing methods has been written, named Dr. Jekyll. The input images to this application is assumed to be segmented before it is invoked.

A survey is given of the design goals of the application, as well as an overview of why the C++ language and a selection of libraries were chosen for the implementation.

Special focus is given to the design and implementation of algorithms in Dr. Jekyll. Effort was made to make the implementation as generic as possible, without sacrificing runtime speed. It is demonstrated how it is possible to use the template mechanism of C++ to achieve this goal.

Classical imaging algorithms, such a connected components analysis and mathematical morphology, is traditionally applied to binary or gray level images. This work formulates a version of these algorithms for segmented images.

Both the application and the algorithms implemented are generic in the sense that they are not tied to a particular field or imaging modality.

Furthermore the application is designed to be extendable, and to provide generic mechanism to implement other image processing tools. To further encourage such extendability the application is licensed under an open source license.