DIGITAL RECONSTRUCTION OF FOSSILIZED ORGANISMS FROM SERIAL IMAGE DATA

The study of organisms inextricably preserved in highy recrystallized rock is often limited by the inaccessability of external and internal morphology. Our work explores new methods for reconstructing detailed three dimensional models of both external and internal structures from serial image data.

We employed new image processing techniques developed for the analysis of medical data to register, segment, and model specimens of tabulate corals and rudistid bivalves from images of parallel cross-sections taken .5mm and 1mm apart, respectively.

To model the coral specimen, we used the Matlab Image Processing Toolkit to register and segment images. We performed a modified version of the thresholding method of image segmentation, combined with binary masking to preserve interior detail. Using Matlab MRI functionality, we created three dimensional models that can be viewed with cut-away slice planes to reveal internal structure.

Our rudistid models were created with a combination of image processing and reconstruction techniques available from the Matlab Image Processing Toolkit, the Insight Toolkit (ITK), and the Visualization Toolkit (VTK). Here we investigated new methods for image segmentation including region growing, level set, and watershed segmentation. We examined Matlab reconstructions versus those of ITK and VTK. Our models include three dimensional reconstructions of entire Hippuritella colonies, detailed models of single individuals, and models cut-away with slice planes taken at arbitrary angles.