Thresholding by Indicator Kriging
IWe consider the problem of segmenting a digitized
image consisting of two univariate populations. Assume a-priori knowledge
allows incomplete assignment of voxels in the image, in the sense that
a fraction of the voxels can be identified as belonging to population
P0, a second fraction to P1, and the remaining fraction have no a-priori
identification. Based upon estimates of the short length scale spatial
covariance of the image, we have developed a method utilizing indicator
kriging to complete the image segmentation.
This work was motivated by three dimensional synchrotron X-ray computed
tomographic (CAT) or laser scanning confocal microscopic (LSCM) images
of biphase materials, such as rock samples, which for our purposes consist
of a material (object) and a void (background) phase.
Identifying the shape of the object is complicated by the partial voxel
(finite volume of resolution) effect as well as by other noise due to
tomographic reconstruction or data quality.