Derived Fields - Scientific Visualization: Uncertainty, Multifield, Biomedical, and Scalable Visualization - page 132

Fig. 14.1 Tracking flow features by treating a stack of vector fields as spin images and using vector
and tensor field measurements [ 27 ]. © IEEE Reprinted, with permission, from IEEE Conference
on Computer Vision and Pattern Recognition, 2009
Xu et al. [ 27 ] track features in a flow dataset by treating the 3D flow field as a
stack of 2D vector fields. Given a point in one slice in the stack, points in other slices
can be correlated to this point by treating the 2D vector fields as spin images. This
idea is then extended to using the velocity gradient tensor fields of the vector fields,
leading to more efficient feature matching (Fig. 14.1 ).
14.2.2 Gradient Comparison
Correlation between a pair of scalar fields has also been defined based on the gradi-
ents. The use of the gradients allows the incorporation of the spatial locality into the
correlation computation. We now describe two derived fields that compare gradients
and discuss their relative merits.