Curvature estimation from orientation fields
M. Van Ginkel, Joost van de Weijer, Lucas J. van Vliet, P.W. Verbeek · Research Repository (Delft University of Technology) · 1999
We present the application of a local dimensionality estimator to the analysis of 3-D microscopic network structures.Three-dimensional images of these structures have been acquired with a fluorescence confocal microscope.We derive the smoothed gradient square tensor (GST) in 3D and show how the eigenvalues and eigenvectors of the tensor can be computed analytically.The eigenvalues yield the dimensionality, the eigenvectors the corresponding orientation.The application of the GST to analyse isotropic, cylindrical and planar structures is tested on synthetic data.The GST analysis of the confocal data requires a reliable measurement of the fluorescence intensities as well as a adequate resolution.We shown that an optimisation of the fluorescence staining combined with an attenuation correction guarantees the former, whereas image restoration will deliver the latter.Finally, results of the application of the GST to confocal data are presented.