denoise.dwt.2d: Denoise an Image via the 2D Discrete Wavelet Transform

View source: R/two_D.R

denoise.dwt.2dR Documentation

Denoise an Image via the 2D Discrete Wavelet Transform

Description

Perform simple de-noising of an image using the two-dimensional discrete wavelet transform.

Usage

denoise.dwt.2d(
  x,
  wf = "la8",
  J = 4,
  method = "universal",
  H = 0.5,
  noise.dir = 3,
  rule = "hard"
)

Arguments

x

input matrix (image)

wf

name of the wavelet filter to use in the decomposition

J

depth of the decomposition, must be a number less than or equal to log(minM,N,2)

method

character string describing the threshold applied, only "universal" and "long-memory" are currently implemented

H

self-similarity or Hurst parameter to indicate spectral scaling, white noise is 0.5

noise.dir

number of directions to estimate background noise standard deviation, the default is 3 which produces a unique estimate of the background noise for each spatial direction

rule

either a "hard" or "soft" thresholding rule may be used

Details

See Thresholding.

Value

Image of the same dimension as the original but with high-freqency fluctuations removed.

Author(s)

B. Whitcher

References

See Thresholding for references concerning de-noising in one dimension.

See Also

Thresholding

Examples


## Xbox image
data(xbox)
n <- nrow(xbox)
xbox.noise <- xbox + matrix(rnorm(n*n, sd=.15), n, n)
par(mfrow=c(2,2), cex=.8, pty="s")
image(xbox.noise, col=rainbow(128), main="Original Image")
image(denoise.dwt.2d(xbox.noise, wf="haar"), col=rainbow(128),
      zlim=range(xbox.noise), main="Denoised image")
image(xbox.noise - denoise.dwt.2d(xbox.noise, wf="haar"), col=rainbow(128),
      zlim=range(xbox.noise), main="Residual image")

## Daubechies image
data(dau)
n <- nrow(dau)
dau.noise <- dau + matrix(rnorm(n*n, sd=10), n, n)
par(mfrow=c(2,2), cex=.8, pty="s")
image(dau.noise, col=rainbow(128), main="Original Image")
dau.denoise <- denoise.modwt.2d(dau.noise, wf="d4", rule="soft")
image(dau.denoise, col=rainbow(128), zlim=range(dau.noise),
      main="Denoised image")
image(dau.noise - dau.denoise, col=rainbow(128), main="Residual image")


waveslim documentation built on June 22, 2024, 9:43 a.m.