NonLinearSmoothArray: Non-linear spatial smmothing of 3D and 4D arrays.

Description

Smooths the values in an array spatially using a weighting kernel that doesn't smooth across boundaries.

Usage

NonLinearSmoothArray(x, voxdim=c(1, 1, 1), radius=2, sm=3, mask=NULL)

Arguments

The array to be smoothed.

voxdim

The voxel dimensions of the array.

radius

The radius of the spatial smoothing

The standard deviation of the Gaussian smoothing kernel.

mask

Optional mask for smoothing.

Value

The smoothed array is returned.

Details

For a 3D array the smoothed values are obtained through a weighted sum of the surrounding voxel values within the specfied radius. The weights are calculated using a Gaussian kernel function applied to the differences between the voxel and its surrounding voxels. In this way the smoothing is anisotropic.

For a 4D array the first 3 dimensions represent space and the fourth represents time. Therefore, each spatial location contains a time series of values. These time series are smoothed spatially in an anisotropic fashion. The sum of squared differences between each pair of time series are used to define the smoothing weights.

Examples

Run this code

#3D array
d<-rep(10,3)
a<-array(3,dim=d)
a[,5:10,5:10]<-7
a<-a+array(rnorm(n=1000,sd=1),dim=d)

h<-NonLinearSmoothArray(a,voxdim=c(1,1,1),radius=2,sm=3)

par(mfrow=c(2,2))
image(a[1,,],zlim=c(-1,12));title("Before smoothing")
image(h[1,,],zlim=c(-1,12));title("After smoothing")
persp(a[1,,],zlim=c(-1,12))
persp(h[1,,],zlim=c(-1,12))

#4D array
d<-c(10,10,10,20)
a<-array(1,dim=d)
a[,,6:10,]<-2
a<-a+array(rnorm(20000,sd=.1),dim=d)

h<-NonLinearSmoothArray(a,voxdim=c(1,1,1),radius=2,sm=3)

par(mfrow=c(2,2),mar=c(0,0,0,0))
for(i in 1:10){
    for(j in 10:1){
        plot(a[1,i,j,],type="l",ylim=c(0,3),axes=FALSE);box()
        lines(h[1,i,j,],col=2)
    }}