R/fft.convolve.R

In AngusWright/helpRfuncs: Helpful functions for R

fft.convolve <-
function(arr1, arr2, normalise=FALSE, mod.arr1=FALSE, renorm.arr2=FALSE, zapdig=NULL) {
  #Details {{{
  # Perform the Convolution of arr1 with arr2 using FFT If mod.arr1 is TRUE the
  # convolution ignores all phase the first image (arr1) - all positional
  # information is determined by the secondary image.  If renorm.arr2 is TRUE,
  # then all magnitude information is determined by the first array If
  # normalise is true, then the final Fourier arrays are normalised.  Use of
  # normalise allows for additional error handling by
  # scaling the zapsmall function to the degree of overflow in the final
  # aperture. I.e. digits = 4 means the final image integral is truncated at 
  # 10^-4 %
  # }}}
  #If wanted, remove array 2 magnitude information {{{
  if (renorm.arr2) {
     #Disregard Magnitude information
     #by normalisation
     arr2<-arr2*sum(arr2)/length(arr2)
  }#}}}
  #Perform Convolution {{{
  if (mod.arr1) {
    step<-Mod(fft(arr1))*fft(arr2)
  } else { 
    step<-fft(arr1)*fft(arr2)
  }
  conv<-Re(fft(step,inverse=TRUE))
  #}}}
  #Normalise convolved array {{{
  if (normalise) {
    conv=conv/length(step)
  }#}}}
  #If needed, get the default zapdigit {{{
  if (is.null(zapdig)) { zapdig<-getOption("digits") }
  #}}}
  # 'zap' small values to avoid numeric errors from fft, and return {{{
  return=zapsmall(conv, digits=zapdig)
  #}}}
}