Sie sind auf Seite 1von 58

Introduction

Noise pdf modeling and estimation


Signal pdf modeling
Conventional noise reduction methods
Gaussian noise
Filtering mean filter, wiener filter, bilateral filter
Thresholding
Bayesian estimation
Block matching 3D (BM3D) image denoising
Poisson-Gaussian noise
Variance Stabilizing Transform (VST)
Poisson-Gaussian unbiased risk estimate-linear
expansion transform (PURE-LET)
Poisson-Gaussian noise reduction using the hidden Markov
modeling with contourlet transform for fluorescence
microscopy images
Hidden Markov Models (HMM)
New transforms: curvelet and contourlet transforms
Poisson-Gaussian contourlet HMM for noise reduction
Experimental results
Conclusion

Observed image data y is different from the original


signal x. It is corrupted by unwanted noise .
y (i ) x(i ) (i )

i : pixel position

Noise comes from thermal noise of analog circuits, photon


counting statistics, or quantization.
Noise strength may depend on the signal.
Noise is affected by nonlinearity of the image acquisition
process, such as clipping or saturation.

Goal
Remove noise while preserving the useful information.

Motivation
Preprocessing or joint processing for super resolution
imaging, image segmentation or quantitative analysis.
Construct an optimal estimation of the original signal based
on the observed image and the noise model.

Gaussian Noise
Noise is independent of signal with Gaussian pdf.

Poisson Noise
Noise variance is proportional to the signal.
Photon counting process is modeled with Poisson distribution.
When signal is large enough, the Poisson probability distribution
can be approximated as Gaussian pdf.

Mixture of Poisson and Gaussian noise


Practical noise model for a camera: Mixture of Poisson and Gaussian
noise.
Poissonian part: modeling of the photon sensing.
-The Poissonian noise variance is proportional to the signal.

Gaussian part: Remaining stationary disturbances in the output data.


-The Gaussian noise variance is independent of the signal.

[Foi 2008]

The need for new noise modeling due to photon noise which is
signal-dependent.

z ( x) y ( x) p y ( x) g ( x)
Poisson signal-dependent
component

Gaussian signalindependent component

x X : the pixel position in the domain X,


z : the observed signal, y : the original signal

1
y ( x) p y ( x)
a

P y ( x) , g ( x)
a

N (0, b) (a 0, b 0)

P( ) : Poission r.v. N( ) : Gaussian r.v.


[Foi 2008]

1
1
E y ( x) p y ( x) var y ( x) p y ( x) y ( x)
a

a
a
1
1
1
1
E p y ( x) 0
E y ( x) p y ( x) y ( x) E p y ( x)
a
a
a
a

1
1
var

y
(
x
)

y ( x)

p

a
a

var p y ( x) ay ( x), var g x b

2 y( x) ay( x) b
Poisson signal-dependent
component

Gaussian signalindependent component [Foi 2008]

Wavelet transform

Segmentation of the input image


into smooth regions using level set

Local estimation for each uniform


regions using wavelet analysis

Global parameter estimation


of a and b
[Foi 2008]

LS =0.0016667 nsteps =385 icounter =385 LS median=57 =129.3192 a=0.0077328

b=0.0034931

0.12

0.1

0.08

0.06

0.04

0.02

Observed image

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Least-squares initialization of the


standard derivation function
initial a: 0.0077, initial b: 0.0035
ML MAD a=0.0097967 b=0.0020081 -logL=-1392.4138
0.12

0.1

0.08

0.06

0.04

0.02

Smoothed approximation
coefficients

Edge removed
detail coefficients

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Maximum likelihood(ML) estimates of


the standard derivation function
a: 0.0098, b: 0.0020
conditional densities ( [0,1] uniform prior density )

210000

10
20
30

800

40
50
60
70
80

0.08

Comparison for noise parameter estimation

Simulation noise

0.5

parameters

400

225

100

25

estimated a

-0.09

0.38

1.08

2.96

5.21

estimated b

430.94

255.88

101.71

47.37

estimated a

0.19

1.23

1.92

3.93

5.55

estimated b

464.45

246.08

101.16

8.70

estimated a

0.03

0.41

1.01

2.95

4.97

estimated b

404.34

240.91

101.93

25.00

0.32

Lena

Barbara

Camera man

Generalized Gaussian model: px ( x) exp( x / s )


The exponent p controls the shape, especially tail part of the pdf.
Laplacian: p =1, Gaussian: p =2
0.5< p <0.8 for natural image wavelet coefficients
Hard to solve
d

Gaussian Scale Mixture model: x

zu

: equality in distribution

x : GSM random vector, u : zero mean Gaussian random vector,


z : a positive scalar random variable
r.v. z and u are independent
GSM family includes Cauchy distribution, generalized Gaussian,
symmetrized Gamma distribution.
Pdf of x is a gaussian, when conditioned on z.
exp
px (x)

p(x | z ) pz ( z )dz

x T ( zC u ) 1 x
2

(2 ) N /2 zCu

1/2

[Portilla 2003]
pz ( z )dz

Independent mixture model with a hidden state


model
employed to match the non-Gaussian nature of the
wavelet coefficients
1) a discrete random state variable, S, having the values
s 1,..., M with pdf pS ( s)
2) The hidden state can represent the interrelationships
of neighboring variables.
3) the conditional Gaussian pdf of W, pW |S (w | S s), s 1,..., M .
the pdf of W is a sum of weighted pW | S
pW ( w)

S (m) pW |S ( w | S

m 1

m).

[Crouse 1998]

Traditional Approach spatial averaging

Observation
Signal has high correlation in space domain
Noise components are independent

Solution
Spatial averaging

Traditional Approach spatial averaging

Spatial Averaging (cont.)


Mean-filtering
v(m, n) a(k , l ) y (m k , n l )
( k .l ) W

, M k M , N l N

a(k , l ) (2M 1)(2 N 1)

0,
elsewhere

y(m, n) u(m, n) (m, n)

(m, n) ~ N (0, 2 )

Noise reduction
v(m, n)

1
u (m k , n l ) (m, n)

N w ( k ,l ) W

(m, n) ~ N (0, ), where


2

Nw

a(k , l )

1
Nw

Traditional Approach spatial averaging

Spatial Averaging (cont.)


Examples of spatial averaging masks a(k,l)
l

0
1

1
4
1
4

1
4
1
4

0
1

(a) 2 x 2 window

1
9
1
9
1
9

1
9
1
9
1
9

1
9
1
9
1
9

(b) 3 x 3 window

1 0

1
8

1 0

0
1
8
1
2
1
8

1
0
1
8

(c) 5-point weighted


averaging

Spatial averaging masks a(k, l)


Note: The sum of the elements is ONE.

Disadvantage : blurring, windowing effect


17

Traditional Approach spatial averaging

Observation
Signal energy has much energy in low frequency.
Noise components have much energy in high
frequency range.

Solution
Filtering considering the noise and signal
spectrum

The Wiener filter restores images in the


presence of degradation and noise.
Results in minimum mean square error
estimates
G f1 , f 2

H * f1 , f 2
H f1 , f 2

1/[ Suu f1 , f 2 / S

f1 , f 2 ]

H f1 , f 2 : Image degradation
Suu f1 , f 2 / S

G f1 , f 2
when H

f1 , f 2 : Signal to Noise ratio

1 1/[ Suu

1
f1 , f 2 / S

1, i.e. without blurring.

f1 , f 2 ]

BF is a Shift-Variant Filter defined by the following equation.


This filtering is simple with high performance .
f [m, n] h[m, n; k , l ]g[k , l ]
k

f [m, n] : recovered image, h[m, n; k , l ] : impulse response at [m, n] from an impulse at [k , l ]


g[m, n] : observed image

Filtering function for BF

Domain filter
(Gaussian LPF)

Range filter

1
(m m0 ) 2 (n n0 ) 2
g[m, n] g[m0 , n0 ]
r
exp

exp

, [m, n] m0 , n0
2
2
h[m0 , n0 ; m, n] m0 , n0
2

d
r

, else
0
[m0 , n0 ] : center position for filtering
m0 , n0 : filtering window centered at [m0 , n0 ]

: Parameter for domain filter

r : parameter for range filter

[Tomasi 1998]

Gaussian LPF (Domain filter)

Bilateral filter (Domain filter* Range filter)

p
pixel
intensity

q
domain

[Durand 02]

domain

pixel position

GB [ I ]p G || p q || I q
qS

domain

BF [ I ]p

1
Wp

pixel position

G || p q || G | I
qS

normalization

Observed
image

range

pixel
intensity

Gaussian LPF

Observed
image

domain

I q | I q

range

Bilateral filtering

Input image

Output image

*
*
(Domain filter)

[Durand 2002]

Input image

Output image

*
*
(Domain filter* Range filter)
Preserves edges

[Durand 2002]

Hard threshold
If an input is larger than threshold value, it is
maintained. Otherwise, the input is replaced to zero.

1 () I ( T )
w : wavelet coefficients,
T: threshold value
I : indicator function

Since hard threshold is discontinuous, it may


cause artifacts on the denoised image.

Soft threshold
A method decreasing an input toward zero according to
a threshold value

2 () ( sgn()T ) I ( T )
The bias is larger than hard thresholds

Semisoft threshold
Complement the drawbacks of hard and soft thresholds
T2 ( T1 )
3 ( ) sgn( )
I (T1 T2 )
T2 T1
(T2 T1 0)

<Flow>

o vn

y w n'

v : original image
o : noisy image
n : Gaussian random variable

w : original image coefficient


y : noisy image coefficient
n : noise coefficient

Estimate w from observed y

k ( yk ) arg max pw y ( wk yk )
w

The Bayesian denoising method is a basic algorithm using


the maximum a posteriori (MAP) estimator with a
Gaussian noise density.
y w n

Wavelet domain:
Observed image

Gaussian noise

Original image

arg max pw|y (w | y)


MAP estimation: w(y)
w

w(y)
arg max py|w (y | w) pw (w).
w

Bayes rule

w(y)
arg max[
w

(y w)2
2 n2

w2
2 w2

w(y)
arg max[ pn (y w) pw (w)]

]
Logarithm

w2

w(y)
2
y
2
w n

[Mihcak 1999]
[Simoncelli 1999]

BM3D is a image denoising algorithm based on an enhanced


sparse representation in transform domain
The enhancement of the sparsity is achieved by grouping similar
2-D image blocks into 3-D data arrays.
A significant improvement is obtained by a specially developed
collaborative Wiener filtering.

One of the state-of-the-art denoising algorithms

[Dabov 2007]

Poisson noise is manipulated to resemble Gaussian


distribution by stabilizing the variance.
VST is applied to Poisson noise to employ well established
data analysis techniques for constant variance.
Conventional Anscombe transformation
f ( x) x c

x: Poisson distribution for x with mean


c: non-negative constant

The mean and variance of f(x)


1

24c 7

E( f ( x))

var( f ( x))

1 3 8c 32c 2 52c 17

4
8
32 2

1/2

128

3/2

When c=3/8, the


variance becomes
constant
[Anscombe 1948]

Modified Anscombe transformation for Mixed


Poisson-Gaussian distribution
Poisson-Gaussian process y:
y np ng , np ~ P( ), ng ~ N (0, 2 )
f ( y) y c

The mean and variance of f(y)


1

24c 7 16 2

When c=3/8+2, the


variance becomes
constant

E( f ( y ))

var( f ( y ))

1 3 8c 8 2 96c 2 52c 64 2 7

1
.
2
4
8
128

81/2

128 3/2

[Starck 1998]

[Zhang 2007]

Poisson-Gaussian unbiased risk estimate-linear


expansion transform (PURE-LET)
PURE-LET algorithms is a state of the art method to denoise Poisson-G
aussian noise .
A method which performs thresholding in transform domain
To improve the performance: Optimization with unbiased estimation of
MSE and using several basis such as undecimated WT(UWT) and
overcomplete M-block DCT(BDCT)
Noise signal modeling:
y = z + b zn ~ P( xn )
z : a vector of N independent zn
b ~ N (0, 2 I )

Unbiased MSE estimate (PURE)

1
2
( f (y ) 2y T f (y ) 2 2div{f ( y )})
N
1
2
( y 1T y ) 2
N

[Luisier 2011]

1
2
( f (y ) 2y T (f (y ) f (y )))
N
1
2
(2 2 div{f (y ) f (y )} y 1T y ) 2
N

approximation

Poisson-Gaussian unbiased risk estimate-linear


expansion transform (PURE-LET)

Signal dependent threshold value


t j ( w) j w 2
j 2 j /2 for multiscale transform
j M 1/2 for overcomplete BDCT
(M : the size of the considered blocks)

Hidden Markov modeling algorithm is a method which


models the properties of wavelet and so it is used effectively
for denoising.

Properties of wavelet
Clustering: If a wavelet coefficient is large/small, then adjacent
coefficients are very likely to also be large/small (horizontal
dependency)
Persistence: Large/small values of wavelet coefficients tend to
propagate across scales (vertical dependency)

(a) clustering

(b) persistence

[Crouse 1998]

Statistical models to represent the signal prope


rties of wavelet in HMM
Independent mixture model: To match the non-Gaussian
nature of the wavelet coefficients, modeling the marginal
probability of each coefficients as a mixture density with a
hidden state variable
Hidden Markov models
Hidden Markov chain : dependency within scale (horizontal)
Hidden Markov tree: dependency across scale (vertical)

Most wavelet coefficients have small values and a few


wavelet coefficients have large values that represent
significant signal information (non-Gaussian property)

Two-state, zero-mean Gaussian mixture model for a random variable W


S=1 corresponds to a low-variance Gaussian pdf
S=2 corresponds to a high-variance Gaussian pdf

M-state Gaussian mixture model for random variable W


fW ( w)

m 1

(m) fW |S ( w | S m).

HMM parameters using M-state Gaussian mixture model for each


wavelet coefficient Wi

pS1 (m) : the pdf for the root node S1

imr
, (i ) pSi |S ( i ) [m | S (i ) r ] : the conditional probability that Si

is in state m given Sp(i) is in state r

i ,m , i2,m : the mean and variance of the wavelet coefficient Wi


given Si is in state m

To search the HMM parameters that fit the given data best, an
expectation maximization (EM) algorithm is employed.
2
Denoise through Baysian estimator using the parameters pS1 (m), i ,m
2

i
E[ yik | w k , ] p( Sik m | w k , ) 2 ,m 2 wik
n i ,m
m

Tiling of the frequency plane


for compact representation of
curve
localized + directional
the frequency window Uj
defined in the Fourier domain

[1]

Freq. domain support of a curvelet

[Starck 2002]

Curvelet transform is simple in the continuous domain,


however, causes problems during the implementation for
discrete images -> the typical rectangular-sampling grid
imposes a prior geometry to discrete images such as strong
bias toward horizontal and vertical directions
The development of a directional multiresolution
transform similar to curvelet, but directly in the discrete
domain -> contourlet

<wavelet>

[Do 2005]

<contourlet>

<Directional Filter Bank>


(a) frequency division
(b) tree structure DFB

<Laplacian Pyramid> (a) decomposition (b) reconstruction

[Do 2005]

Examples of the frequency partitioning

Example of the contourlet transform on a


test image. The image is decomposed into
two levels, which are then decomposed
into four and eight directional subbands.

Directional filter bank

Noise variance in the contourlet domain

Contourlet coefficient of noisy signal: wi zi k hk zi hi .


Noise variance in the contourlet domain:
n2i E[wi 2 ] E[(

zi k hk )2 ].

The cross product terms of ( zi k hk )2 are eliminated,


since neighboring noise components are assumed to be
statistically independent.

(z

n2i E[

i k

hk 2 )]

(E[ z

i k

]hk 2 )

2
2
zik hk )

z2ik hk2 .

Obtained by filtering estimated noise variances in the


image domain using the square of the contourlet filter
coefficients.

The conditional mean estimation of can be


obtained by using the hidden state probabilities
p(Sik | w k , ) as by-products of the EM algorithm
through the chain rule of conditional expectation.
vik

E[vik

| w , ]
k

p( Sik

i2,m
k
m | w , ) 2
w
i
ni i2,m
k

The denoised signal is acquired by the inverse


transform of the estimated contourlet coefficients.

Cycle spinning method


When x s noise,

1
s
K1K 2
^

K1 , K 2

i , j

i 1, j 1

Si , j

K1 , K2

Transform

Denoising

Inverse

Transform

(T 1 ( [T ( Si , j ( x)]))

w=y+n

<Conventional method>
x

WT

WGN

EM
Algorithm

Bayesian
Estimation

Using HMT
models
w=y + n

<PG-HMM>
x

noise

WT-1

EM
Algorithm

o x P ( x) g

CT

w CT (o)
y CT ( x)
n CT ( p ( x) g )

a, b estimation
for noise
variance

Bayesian
Estimation

Using HMT
models
Cycle spinning

Wiener
filtering

y
CT-1

Cell

Lena

Barbara

10

20

30

40

Noised image (dB)

28.14

22.11

18.59

16.09

HMM based on wavelet (dB)

36.27

32.30

30.00

28.54

HMM based on contourlet (dB)

37.75

33.67

31.39

29.64

Noised image (dB)

28.14

22.11

18.59

16.09

HMM based on wavelet (dB)

33.94

30.48

28.68

27.41

HMM based on contourlet (dB)

35.12

31.84

30.07

28.73

Noised image (dB)

28.14

22.11

18.59

16.09

HMM based on wavelet (dB)

31.53

27.49

25.52

24.05

HMM based on contourlet (dB)

32.49

28.67

26.81

25.22

This table shows that HMM based on contourlet is more effective.


MSEs of the denoised image have been averaged over 10 realizations

(a)

(b)

(c)
(d)
Results on cell image (a) original image (b) noisy image with n=20. Gaussian noise
(22.11 dB) (c) HMM based on wavelet (32.30 dB) (d) HMM based on contourlet (33.67 dB)

Comparison of denoising performance: fixed noise


variance value with various rate of Poisson and
Gaussian (PSNR)
Noise variance
Image

Lena

2.65

5.59

5.78

737.32

202

102

52

Noisy image

19.46

19.46

19.47

19.47

19.45

Gaussian wavelet HMM

29.10

28.95

28.25

27.94

27.79

Gaussian contourlet
HMM

29.45

29.28

28.51

28.22

27.91

Gaussian contourlet
HMM with modified
Anscombe

29.09

28. 68

28.65

29.13

29.10

Proposed mixed PoissonGaussian contourlet


HMM

30.14

30.09

30.04

29.96

30.02

The proposed method shows the robust results regardless of noise rate
of Poisson and Gaussian.
Better performance than the conventional method with Anscombe.

Comparison of denoising performance of the


proposed method and BM3D on low-count images
with various noise (PSNR)
Noise variance
Image

Lena

Camera man

Cell

255

255/2

255/3

255/5

255/10

0.12

0.22

0.32

0.52

12 0

Noisy image

2.97

5.97

7.85

8.58

13.01

BM3D with modified


Anscombe

16.80

22.70

24.65

26.37

27.80

Proposed contourlet HMM


with Bayesian estimation

21.28

22.94

24.72

24.88

25.98

Noisy image

3.26

6.20

8.84

10.85

13.34

BM3D with modified


Anscombe

15.84

20.02

25.57

26.37

28.99

Proposed contourlet HMM


with Bayesian estimation

19.59

22.59

24.77

26.43

28.16

Noisy image

3.22

6.17

7.83

9.84

12.51

BM3D with modified


Anscombe

16.22

20.07

22.13

24.10

26.96

Proposed contourlet HMM


with Bayesian estimation

19.77

21.74

22.84

23.98

26.86

The proposed method shows better performance when the number of


photons is limited.

Original Image

Noisy Image (7.85 dB)

(a) (8.15 dB)Contourlet + Wiener


Contourlet Denoising
(b)Denoising (8.16 dB)

Results on Lena image

(c)

(a) original image

(d)

(b) noisy image with a=1/3, b=0.32 (7.85 dB)

(c)BM3D (24.65 dB) (d) the proposed method (24.72 dB)

(a)

(b)

(c)
(d)
Results on the magnified Lena image
(a) original image (b) noisy image with a=1/3, b=0.32 (c) BM3D (d) the proposed method

Comparison of denoising performance of


fluorescent microscopic image (PSNR)

R channel G channel B channel

total

Noisy image

28.64

21.95

28.41

26.33

Multiscale estimation with


Haar wavelet

34.62

27.05

35.61

32.43

Bilateral filtering

34.35

26.45

36.81

32.54

BM3D with modified


Anscombe

32.69

25.23

39.97

32.63

Proposed contourlet HMM


with Bayesian estimation

35.58

29.53

36.94

34.02

Acquired on a C1 Plus confocal laser scanning microscope from Nikon.


Provided by the Medical Bioconversions Center, Seoul National University.
100 instances of 512x512 images of fixed Hera cells, labeled with
three fluorescent dyes(green: alexafluor 488, red: alexafluor 555, blue: DAPI)
Ground truth: The average of the 100 images.
The proposed method performs better than other existing algorithms.

(a)

(b)

(c)

(d)

(e)

(f)

Results on the magnified fluorescent microscopic image


(a) averaged image of 100 acquisitions (b) noisy image(single acquisition)
(c) multiscale estimation with Haar
(d) Bilateral filtering
(e) BM3D (32.63 dB)
(f) the proposed method (34.32 dB)

To reduce noise we need to know


Noise pdf: Gaussian, Poisson, or mixture of them
Signal model: Generalized gaussian, Gaussian scale
mixture, HMM

Noise reduction techniques


Gaussian noise reduction
Image domain: local averaging, bilateral filter
Frequency domain:
Many frequency transforms: Fourier, Wavelet, contourlet, curvelet, etc.
Thresholding: soft or hard
Wiener filtering

Poisson-Gaussian noise reduction


VST + Gaussian noise reduction methods
PURELET
Contourlet + HMM

1.

2.
3.
4.

5.
6.
7.
8.
9.

A. Foi, M. Trimeche, V. Katkovnik, and K. Egiazarian, Practical Poissonian-Gaussian noise


modeling and fitting for single-image raw-data, IEEE Trans. on Image Processing, vol. 17, no.
10, pp. 1737-1754, Oct. 2008.
F. Luisier, T. Blu, and M. Unser, Image Denoising in Mixed Poisson-Gaussian Noise, IEEE Trans.
on Image Processing, vol. 20, no. 3, pp. 696-708, Mar. 2011.
H. Robbins, The empirical Bayes approach to statistical decision problems, Ann. Math.
Statist., vol. 35, pp. 120, 1964.
J. S. Lee, Digital image enhancement and noise filtering by use of local statistics, IEEE Pattern
Anal. Machine Intell., vol. PAMI-2, pp. 165168, Mar. 1980.
M. Malfait and D. Roose, Wavelet-based image denoising using a Markov random field a
priori model, IEEE Trans. Image Processing, vol. 6, pp. 549565, Apr. 1997.
M. S. Crouse, R. D. Nowak, and R. G. Baraniuk, "Wavelet-based Statistical Signal Processing
Using Hidden Markov Models," IEEE Trans. on Signal Processing, vol. 64, no. 4, pp. 886-902,
April 1998.
E. P. Simoncelli, Bayesian Denoising of Visual Images in the Wavelet Domain, in Bayesian
Inference in Wavelet Based Models, P. Muller and B. Vidakovic, Eds. New-York: Springer-Verlag,
1999, vol. 141, ch. 18, pp. 291-308.
M. K. Mihcak, I. Kozintsev, K. Ramchandran, and P. Moulin Low-Complexity Image Denoising
Based on Statistical Modeling of Wavelet Coefficients, IEEE Trans. on Signal Processing Letters,
vol. 6, no. 12, pp. 300-303, 1999.
J. Portilla, V. Strela, M. J. Wainwright, and E. P. Simoncelli, Image denoising using scale
mixtures of Gaussians in the wavelet domain, IEEE Trans. on Image Processing, vol. 12, no. 11,
pp. 1338-1351, Nov. 2003.

10. K. Dabov, A. Foi, V. Katkovnik, and K. Egiazarian, Image Denoising by Sparse 3-D Transform-Domain
Collaborative Filtering, IEEE Trans. on Image Processing, vol. 16, no. 8, pp. 2080-2095, Aug. 2007.
11. F. J. Anscombe, The Transformation of Poisson, Binomial and Negative-Binomial Data, Biometrika,
vol. 35, no. 3/4, pp. 246-254, Dec. 1948.
12. M. Fisz, The limiting distribution of a function of two independent random variables and its
statistical application, Colloq. Math., vol. 3, pp. 138-146, 1955.
13. D. L. Donoho, Nonlinear Wavelet Methods for Recovery of Signals, Densities, and Spectra from
Indirect and Noisy Data, in Proceedings of Symposia in Applied Mathematics, American
Mathematical Society, pp. 173-205, 1993.
14. P. Fryzlewicz and G. P. Nason, A Haar-Fisz Algorithm for Poisson Intensity Estimation, Journal of
Computational and Graphical Statistics, vol. 13, no. 3, pp. 621-638, 2004.
15. M. Jansen, Multiscale Poisson Data Smoothing, Journal of the Royal Statistical Society B, vol. 68, no.
1, pp. 2748, 2006.
16. B. Zhang, J. M. Fadili, and J.-L. Starck, J.-C. Olivo-Marin, Multiscale Variance-Stabilizing Transform for
Mixed-Poisson-Gaussian Processes and its Applications in Bioimaging, Proc. IEEE Int. Conf. Image Pro
cess., vol. 6, pp. 233-236, 2007.
17. B. Zhang, J. M. Fadili, and J.-L. Starck, Wavelets, Ridgelets, and Curvelets for Poisson Noise Removal,
IEEE Trans. on Image Processing, vol. 17, no. 7, pp. 10931108, 2008.
18. S. Lefkimmiatis, P. Maragos, and G. Papandreou, Bayesian inference on multiscale models for Poisson
intensity estimation: Applications to photon-limited image denoising, IEEE Trans. Image Processing,
vol. 18, no. 8, pp. 1724-1741, 2009.
19. F. Luisier, C. Vonesch, T. Blu, and M. Unser, Fast interscale wavelet denoising of Poisson-corrupted
images, Signal Processing, vol. 90, no. 2, pp. 415-427, 2010.

20. M. Makitalo and A. Foi, Optimal Inversion of the Anscombe Transformation in Low-count Poisson
Image Denoising, IEEE Trans. on Image Processing, vol. 20, no. 1, pp. 99-108, 2011.
21. D. L. Donoho, De-noising by soft-thresholding, IEEE Trans. Inform. Theory, vol. 41, pp. 613627, May
1995.
22. E. D. Kolaczyk, Nonparametric estimation of intensity maps using Haar wavelets and Poisson noise
characteristics, Astrophys. J., vol.534, pp. 490-505, 2000.
23. K. E. Timmermann and R. D. Nowak, Multiscale modeling and estimation of Poisson processes with
application to photon-limited imaging, IEEE Trans. Inf. Theory, vol. 45, no. 2, pp. 846-862, Apr. 1999.
24. E. D. Kolaczyk, Bayesian multi-scale models for Poisson processes, J. Amer. Stat. Assoc., vol. 94, no.
447, pp. 920-933, Sep. 1999.
25. H. Lu, Y. Kim, and J. M. M. Anderson, Improved Poisson intensity estimation: Denoising application
using Poisson data, IEEE Trans. Image Processing, vol. 13, no. 8, pp. 1128-1135, Aug. 2004.
26. S. Lefkimmiatis, P. Maragos, and G. Papandreou, Bayesian inference on multiscale models for Poisson
intensity estimation: Applications to photon-limited image denoising, IEEE Trans. Image Processing,
vol. 18, no. 8, pp. 1724-1741, Aug. 2009.
27. E. D. Kolaczyk, Wavelet shrinkage estimation of certain Poisson intensity signals using corrected
thresholds, Statistica Sinica, vol. 9, no. 1, pp. 119-135, Jan. 1999.
28. C. Charles and J. Rasson, Wavelet denoising of Poisson-distributed data and applications, Comput.
Stat. Data Anal., vol. 43, no. 2, pp. 139-148, 2003.
29. R. D. Nowak and R. G. Baraniuk, Wavelet-domain filtering for photon imaging systems, IEEE Trans.
Image Processing, vol. 8, no. 5, pp. 666-678, May 1999.
30. J.-L. Starck, F. Murtagh, and A. Bijaoui, Image Processing and Data Analysis, Cambridge University Pre
ss, 1998.

31. C. Stein, Estimation of the mean of a multivariate normal distribution, The Annals of Statistics, vol. 9,
pp. 11351151, 1981.
32. M. S. Crouse, R. D. Nowak, and R. G. Baraniuk, "Wavelet-based Statistical Signal Processing Using
Hidden Markov Models," IEEE Trans. on Signal Processing, vol. 64, no. 4, pp. 886-902, April 1998.
33. J. L. Starck, E. J. Cands, and D. L. Donoho, The Curvelet Transform for Image Denoising, IEEE Trans.
on Image Processing, vol. 11, no. 6, pp. 670-684, June 2002.
34. M. N. Do and M. Vetterli, The contourlet transform: An efficient directional multiresolution image
representation, IEEE Trans. on Image Processing, vol. 14, no. 12. pp. 2091-2106, Dec. 2005.
35. P. J. Burt and E. H. Adelson, The Laplacian pyramid as a compact image code, IEEE Trans. on
Communication, vol. 31, no. 4, pp. 532-540, April 1983.
36. D. D. Y. Po and M. N. Do, Directional multiscale modeling of image using the contourlet transform,
IEEE Trans. on Image Processing, vol. 15, no. 6, pp. 1610-1620, June 2006.
37. R. H. Bamberger and M. J. T. Smith, A Filter Bank for the Directional Decomposition of Images:
Theory and Design, IEEE Trans. on Signal Processing, vol. 40, no. 4, pp. 882-893, April 1992.
38. R. Eslami and H. Radha, The Contourlet Transform for Image De-noising Using Cycle Spinning, in
Proc. Asilomar Conference on Signal, Systems and Computer, pp. 1982-1986, November, 2003, Pacific
Grove, CA.
39. R. R. Coifman and D. L. Donoho, Translation Invariant Denoising, in Wavelets and Statistics, Springer
Lecture Notes in Statistics 103, New York, Springer-Verlag, pp. 125-150, 1994.
40. A. L. Cunha, J. Zhou and M. N. Do, The Nonsubsampled Contourlet Transform: Theory, Design, and
Applications, IEEE Trans. on Image Processing, vol. 15, no. 10, pp. 3089-3101, Oct. 2006.

Das könnte Ihnen auch gefallen