boofcv.alg.denoise.wavelet

Class UtilDenoiseWavelet

java.lang.Object

boofcv.alg.denoise.wavelet.UtilDenoiseWavelet

public class UtilDenoiseWavelet
extends java.lang.Object

Various functions useful for denoising wavelet transforms.

Constructor Summary

Constructors

Constructor and Description
UtilDenoiseWavelet()

Method Summary

Modifier and Type	Method and Description
static float	estimateNoiseStdDev(GrayF32 subband, float[] storage) Robust median estimator of the noise standard deviation.
static float[]	subbandAbsVal(GrayF32 subband, float[] coef) Computes the absolute value of each element in the subband image are places it into 'coef'
static double	universalThreshold(ImageGray image, double noiseSigma) Computes the universal threshold defined in [1], which is the threshold used by VisuShrink.

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

UtilDenoiseWavelet

public UtilDenoiseWavelet()

Method Detail

estimateNoiseStdDev

public static float estimateNoiseStdDev(GrayF32 subband,
                                        float[] storage)

Robust median estimator of the noise standard deviation. Typically applied to the HH₁ subband.

σ = Median(|Y_ij|)/0.6745
where σ is the estimated noise standard deviation, and Median(|Y_ij|) is the median absolute value of all the pixels in the subband.

D. L. Donoho and I. M. Johnstone, "Ideal spatial adaption via wavelet shrinkage." Biometrika, vol 81, pp. 425-455, 1994

Parameters:

subband - The subband the image is being computed from. Not modified.

storage - Used to temporarily store the absolute value of each element in the subband.

Returns:

estimated noise variance.

subbandAbsVal

public static float[] subbandAbsVal(GrayF32 subband,
                                    float[] coef)

Computes the absolute value of each element in the subband image are places it into 'coef'

universalThreshold

public static double universalThreshold(ImageGray image,
                                        double noiseSigma)

Computes the universal threshold defined in [1], which is the threshold used by VisuShrink. The same threshold is used by other algorithms.

threshold = σ sqrt( 2*log(max(w,h))
where (w,h) is the image's width and height.

[1] D. L. Donoho and I. M. Johnstone, "Ideal spatial adaption via wavelet shrinkage." Biometrika, vol 81, pp. 425-455, 1994

Parameters:

image - Input image. Only the width and height are used in computing this thresold.

noiseSigma - Estimated noise sigma.

Returns:

universal threshold.