Morph an array with a kernel

The morph function applies a kernel to a target array. Optionally, applying the kernel to a particular array element can be made conditional on its value, or the number of nonzero immediate neighbours that it has. The morph function is (S3) generic.

morph(x, kernel, ...)

# S3 method for default
morph(x, kernel, operator = c("+", "-", "*", "i", "1", "0",
  "=="), merge = c("sum", "min", "max", "mean", "median", "all", "any"),
  value = NULL, valueNot = NULL, nNeighbours = NULL,
  nNeighboursNot = NULL, renormalise = TRUE, ...)

Arguments

x	Any object. For the default method, this must be coercible to an array.
kernel	An object representing the kernel to be applied, which must be coercible to an array. It must have odd width in all dimensions, but does not have to be isotropic in size. The kernel's dimensionality may be less than that of the target array, x. See kernels for kernel-generating functions.
...	Additional arguments to methods.
operator	The operator applied elementwise within the kernel, as a function of the original image value and the kernel value. Arithmetic operators are as usual; "i" is the identity operator, where every value within the kernel will be included as-is; "1" and "0" include a 1 or 0 for each element within the kernel's nonzero region; "==" produces a 1 where the image matches the kernel, and 0 elsewhere.
merge	The operator applied to combine the elements into a final value for the centre pixel. All have their usual meanings.
value	An optional vector of values in the target array for which to apply the kernel. Takes priority over valueNot if both are specified.
valueNot	An optional vector of values in the target array for which not to apply the kernel.
nNeighbours	An optional numeric vector giving allowable numbers of nonzero neighbours (including diagonal neighbours) for array elements where the kernel will be applied. Takes priority over nNeighboursNot if both are specified.
nNeighboursNot	An optional numeric vector giving nonallowable numbers of nonzero neighbours (including diagonal neighbours) for array elements where the kernel will be applied.
renormalise	If TRUE, the default, and merge is "sum", the sum will be renormalised relative to the sum over the visited part of the kernel. This avoids low-intensity bands around the edges of a morphed image.

Value

A morphed array with the same dimensions as the original array.

See also

kernels for kernel-generating functions, and morphology for more specific mathematical morphology functions. gameOfLife shows how this function can be used for non-morphological purposes, in that case to power a cellular automaton. See also the kernel and kernapply functions in the stats package, particularly if you want to smooth time series.

Author

Jon Clayden <code@clayden.org>