imagine (version 2.1.2)

meanFilter: Make a 2D filter calculations from numeric matrix

Description

This functions take a matrix object, and for each cell calculate mean, median or certain quantile around a squared/rectangular neighborhood.

Usage

meanFilter(X, radius, times = 1, na_only = FALSE)
quantileFilter(X, radius, probs, times = 1, na_only = FALSE)
medianFilter(X, radius, times = 1, na_only = FALSE)

Value

A matrix object with the same dimensions of X.

quantileFilter don't use a kernel but, for each cell, it returns the position of quantile 'probs' (value between 0 and 1).

medianFilter is a wrapper of quantileFilter with probs = 0.5.

Arguments

X

A numeric matrix object used for apply filters.

radius

Size of squared or rectangular kernel to apply median. See Details.

times

How many times do you want to apply the filter?

na_only

logical, Do you want to apply the filter ONLY in cells with NA?

probs

numeric vector of probabilities with values in [0,1].

Details

radius must be defined as a 2-length numeric vector specifying the number of rows and columns of the window which will be used to make calculations. If the length of radius is 1, the window will be a square.

Functions use C++ algorithms for running some statistical calculations. The mean is far obvious, however, there are several ways to perform quantiles. quantileFilter function uses arma::quantile: a RcppArmadillo function, which is equivalent to use R quantile funtion with type = 5.

medianFilter is a wraper of quantileFilter, so the same observations are applied to it.

na_only performs two actions at once: (1) it applies the filter only in the positions where the original value is NA and (2) for the rest of the cells, it is replaced with the value of the original matrix.

Examples

# Generate example matrix
nRows <- 50
nCols <- 100

myMatrix <- matrix(runif(nRows*nCols, 0, 100), nrow = nRows, ncol = nCols)
radius <- 3

# Make convolution
myOutput1 <- meanFilter(X = myMatrix, radius = radius)
myOutput2 <- quantileFilter(X = myMatrix, radius = radius, probs = 0.1)
myOutput3 <- medianFilter(X = myMatrix, radius = radius)

# Plot results
par(mfrow = c(2, 2))
image(myMatrix, zlim = c(0, 100), title = "Original")
image(myOutput1, zlim = c(0, 100), title = "meanFilter")
image(myOutput2, zlim = c(0, 100), title = "quantileFilter")
image(myOutput3, zlim = c(0, 100), title = "medianFilter")