6.3. Sample geometry
Irregular sample geometry poses another challenge to
CI. In the case of a non-flat sample, signal differences
may arise from changes in relative path length from the
detector to areas on the sample with variable height.
Non-destructive correction of such morphologicallyinduced
spectral variability may be achieved by developing
post-processing algorithms which are typically application
specific, depending on the geometry of the sample under
consideration; for example, Wang and Paliwal [63] developed
a morphological shrinking algorithm to compensate
for curvature-induced variability in spectral images of
wheat kernels.