Cog 4.0 switched the standard kernel representation from a Matrix to a ScalarField.
Cog 4.0 switched the standard kernel representation from a Matrix to a ScalarField. Old code still uses the Matrix standard. This implicit is here to simplify porting for difficult cases.
-Ben
Implements a polynomial filter bank.
The "input" scalar field is projected locally onto a 2nd-order polynomial approximation, represented by a matrix field + vector field + scalar field. The basis set consists of the following 2-dimensional functions: 1, x, y, x*x, x*y, y*y.
Note that this filter bank introduces no delay.
This is from Gunnar Farneback's dissertation.
use libcog.filters.spatial.QuadraticExpansion.apply() instead