The inverse kernel is the focal filter applied to the coarse raster
before standard interpolation; it predistorts the coarse data so that
subsequent disagg produces a fine raster whose block means equal the
input coarse values.
Usage
kernel(method, fact, radius = NULL)
Arguments
- method
Character, prefilter method name. Currently `"bilinear"`.
- fact
Integer disagg factor.
- radius
Integer half-width of the kernel. Larger = more accurate
at higher one-time cost. `NULL` uses a method-specific default.
Value
A square matrix of side `2 * radius + 1`, with attributes
`method`, `fact`, `radius`, `tail_max` (largest absolute value at
the kernel boundary; an upper bound on truncation error), and
`conv_max_err` (largest deviation of K * K_inv from the delta on
the interior; should be machine epsilon).
Details
Kernels depend only on the method and disagg factor (not the data) and
are cached for the R session.