# Wavelet Block Thresholding¶


This numerical tour presents block thresholding methods, that makes use of the structure of wavelet coefficients of natural images to perform denoising. Theoretical properties of block thresholding were investigated in CaiSilv Cai99 HallKerkPic99

In [2]:
options(warn=-1) # turns off warnings, to turn on: "options(warn=0)"

library(imager)
library(png)

for (f in list.files(path="nt_toolbox/toolbox_general/", pattern="*.R")) {
source(paste("nt_toolbox/toolbox_general/", f, sep=""))
}

for (f in list.files(path="nt_toolbox/toolbox_signal/", pattern="*.R")) {
source(paste("nt_toolbox/toolbox_signal/", f, sep=""))
}
options(repr.plot.width=4, repr.plot.height=4)


## Generating a Noisy Image¶

Here we use an additive Gaussian noise.

Size of the image of $N=n \times n$ pixels.

In [2]:
n <- 256


First we load an image $f_0 \in \RR^N$.

In [3]:
f0 <- rescale(load_image("nt_toolbox/data/boat.png", n))


Display it.

In [4]:
imageplot(f0)


Noise level.

In [5]:
sigma <- .08


Generate a noisy image $f=f_0+\epsilon$ where $\epsilon \sim \Nn(0,\si^2\text{Id}_N)$.

In [6]:
f <- f0 + sigma*as.cimg(rnorm(n**2))


Display it.

In [7]:
imageplot(clamp(f))


## Orthogonal Wavelet Thresholding¶

We first consider the traditional wavelet thresholding method.

Parameters for the orthogonal wavelet transform.

In [8]:
Jmin <- 4


Shortcuts for the foward and backward wavelet transforms.

In [9]:
wav  <- function(f){ perform_wavelet_transf(f, Jmin, +1) }
iwav <- function(fw){ perform_wavelet_transf(fw, Jmin, -1) }


Display the original set of noisy coefficients.

In [10]:
options(repr.plot.width=6, repr.plot.height=6)

plot_wavelet(wav(f), Jmin)


Denoting $\Ww$ and $\Ww^*$ the forward and backward wavelet transform, wavelet thresholding $\tilde f$ is defined as

$$\tilde f = \Ww^* \circ \theta_T \circ \Ww(f)$$

where $T>0$ is the threshold, that should be adapted to the noise level.

The thresholding operator is applied component-wise

$$\th_T(x)_i = \psi_T(x_i) x_i$$

where $\psi_T$ is an atenuation fonction. In this tour, we use the James Stein (JS) attenuation:

$$\psi_T(s) = \max\pa{ 0, 1-\frac{T^2}{s^2} }$$
In [11]:
psi <- function (s,T){
d <- dim(s)
s0 <- 1 - T**2/vapply(as.vector(abs(s)**2), function(a){max(a,1e-9)}, double(1))
s0 <- vapply(as.vector(s0), function(a){max(a,0)}, double(1))
if (max(d,0)!=0){s0 <- array(s0, d)}
return(s0)
}


Display the thresholding function $\th_T$.

In [12]:
options(repr.plot.width=4, repr.plot.height=4)

s <- seq(-3, 3, length=1024)

plot(s, s*psi(s,1), "l", col="blue")
lines(s, s, lty=2, col="red")


Thresholding operator.

In [13]:
theta <- function(x,T){ psi(x, T)*x }
ThreshWav <- function(f,T){ iwav(theta(wav(f), T)) }


Test the thresholding.

In [14]:
T <- 1.5*sigma

options(repr.plot.width=4, repr.plot.height=4)
imageplot(clamp(ThreshWav(f, T)))


Exercise 1

Display the evolution of the denoising SNR when $T$ varies. Store in $f_{Thresh}$ the optimal denoising result.

In [15]:
options(repr.plot.width=4, repr.plot.height=4)

source("nt_solutions/denoisingwav_4_block/exo1.R")

In [16]:
## Insert your code here.


Display the optimal thresolding.

In [17]:
imageplot(clamp(fThresh), paste("SNR = ", round(snr(f0, fThresh),1), "dB"))


## Block Thresholding Operator¶

A block thresholding operator of coefficients $x=(x_i)_{i=1}^P \in \RR^P$ is defined using a partition $B$ into a set of blocks $b$

$$\{1,\ldots,P\} = \bigcup_{b \in B} b.$$

$$\forall i \in b, \quad \theta_T(x)_i = \psi_T\left( \norm{x_b}_2 \right) x_i$$

where $x_b = (x_j)_{j \in B} \in \RR^{\abs{b}}$. One thus thresholds the $\ell^2$ norm (the energy) of each block rather than each coefficient independently.

For image-based thresholding, we use a partition in square blocks of equal size $w \times w$.

The block size $w$.

In [18]:
w <- 4
n <- 256


Compute indexing of the blocks.

In [19]:
source("nt_toolbox/toolbox_wavelet_meshes/meshgrid.R")
grid <- meshgrid_4d(seq(1, n-w+1, w), seq(1, n-w+1, w), seq(0, w-1), seq(0, w-1))
X <- grid$X ; Y <- grid$Y ; dX <- grid$Z ; dY <- grid$S
I <- (X + dX-1) + (Y + dY-1)*n

for (k in 1:(n%/%w)){
for (l in 1:(n%/%w)){
I[k,l,,] <- t(I[k,l,,])
}
}


Block extraction operator. It returns the set $\{x_b\}_{b \in B}$ of block-partitioned coefficients.

In [20]:
block <- function(x){ array(as.vector(x)[I+1], dim(I)) }


Block reconstruction operator.

In [21]:
assign <- function(M,I,H){
M_temp <- as.array(M, dim(I))
M_temp[I+1] <- H
return(matrix(M_temp, c(n,n), byrow=FALSE))
}

iblock <- function(H){ assign(matrix(rep(0,n*n), c(n,n)), I, H)}


Check that block extraction / reconstruction gives perfect reconstruction.

In [22]:
print( paste("Should be 0:", norm(as.matrix(f) - iblock(block(f)))) )

[1] "Should be 0: 0"


Compute the average energy of each block, and duplicate.

In [23]:
energy <- function(H){
H_tmp <- H
for (i in 1:n%/%w){
for (j in 1:n%/%w){
H_tmp[i,j,,] <- sqrt( mean(H_tmp[i,j,,]**2) )
}
}
return(H_tmp)
}


Block thresholding operator.

In [24]:
Thresh <- function(H,T){ psi(energy(H), T)*H }
ThreshBlock <- function(x,T){ iblock(Thresh(block(x), T)) }


Exercise 2

Test the effect of block thresholding on the image $f_0$ itself, for increasing value of $T$. Of course directly thresholding the image has no interest, this is just to vizualize the effect.

In [25]:
options(repr.plot.width=8, repr.plot.height=8)

source("nt_solutions/denoisingwav_4_block/exo2.R")