In statistics, the Anscombe transform, named after Francis Anscombe, is a variance-stabilizing transformation that transforms a random variable with a Poisson distribution into one with an approximately standard Gaussian distribution. The Anscombe transform is widely used in photon-limited imaging (astronomy, X-ray) where images naturally follow the Poisson law. The Anscombe transform is usually used to pre-process the data in order to make the standard deviation approximately constant. Then denoising algorithms designed for the framework of additive white Gaussian noise are used; the final estimate is then obtained by applying an inverse Anscombe transformation to the denoised data.
Definition
For the Poisson distribution the mean
m
{\displaystyle m}
and variance
v
{\displaystyle v}
are not independent:
m
=
v
{\displaystyle m=v}
. The Anscombe transform
A
:
x
↦
2
x
+
3
8
{\displaystyle A:x\mapsto 2{\sqrt {x+{\tfrac {3}{8}}}}\,}
aims at transforming the data so that the variance is set approximately 1 for large enough mean; for mean zero, the variance is still zero.
It transforms Poissonian data
x
{\displaystyle x}
(with mean
m
{\displaystyle m}
) to approximately Gaussian data of mean
2
m
+
3
8
−
1
4
m
1
/
2
+
O
(
1
m
3
/
2
)
{\displaystyle 2{\sqrt {m+{\tfrac {3}{8}}}}-{\tfrac {1}{4\,m^{1/2}}}+O\left({\tfrac {1}{m^{3/2}}}\right)}
and standard deviation
1
+
O
(
1
m
2
)
{\displaystyle 1+O\left({\tfrac {1}{m^{2}}}\right)}
.
This approximation gets more accurate for larger
m
{\displaystyle m}
, as can be also seen in the figure.
For a transformed variable of the form
2
x
+
c
{\displaystyle 2{\sqrt {x+c}}}
, the expression for the variance has an additional term
3
8
−
c
m
{\displaystyle {\frac {{\tfrac {3}{8}}-c}{m}}}
; it is reduced to zero at
c
=
3
8
{\displaystyle c={\tfrac {3}{8}}}
, which is exactly the reason why this value was picked.
Inversion
When the Anscombe transform is used in denoising (i.e. when the goal is to obtain from
x
{\displaystyle x}
an estimate of
m
{\displaystyle m}
), its inverse transform is also needed
in order to return the variance-stabilized and denoised data
y
{\displaystyle y}
to the original range.
Applying the algebraic inverse
A
−
1
:
y
↦
(
y
2
)
2
−
3
8
{\displaystyle A^{-1}:y\mapsto \left({\frac {y}{2}}\right)^{2}-{\frac {3}{8}}}
usually introduces undesired bias to the estimate of the mean
m
{\displaystyle m}
, because the forward square-root
transform is not linear. Sometimes using the asymptotically unbiased inverse
y
↦
(
y
2
)
2
−
1
8
{\displaystyle y\mapsto \left({\frac {y}{2}}\right)^{2}-{\frac {1}{8}}}
mitigates the issue of bias, but this is not the case in photon-limited imaging, for which
the exact unbiased inverse given by the implicit mapping
m
x
e
−
m
x
!
)
↦
m
{\displaystyle \operatorname {E} \left[2{\sqrt {x+{\tfrac {3}{8}}}}\mid m\right]=2\sum _{x=0}^{+\infty }\left({\sqrt {x+{\tfrac {3}{8}}}}\cdot {\frac {m^{x}e^{-m}}{x!}}\right)\mapsto m}
should be used. A closed-form approximation of this exact unbiased inverse is
y
↦
1
4
y
2
−
1
8
+
1
4
3
2
y
−
1
−
11
8
y
−
2
+
5
8
3
2
y
−
3
.
{\displaystyle y\mapsto {\frac {1}{4}}y^{2}-{\frac {1}{8}}+{\frac {1}{4}}{\sqrt {\frac {3}{2}}}y^{-1}-{\frac {11}{8}}y^{-2}+{\frac {5}{8}}{\sqrt {\frac {3}{2}}}y^{-3}.}
Alternatives
There are many other possible variance-stabilizing transformations for the Poisson distribution. Bar-Lev and Enis report a family of such transformations which includes the Anscombe transform. Another member of the family is the Freeman-Tukey transformation
A
:
x
↦
x
+
1
+
x
.
{\displaystyle A:x\mapsto {\sqrt {x+1}}+{\sqrt {x}}.\,}
A simplified transformation, obtained as the primitive of the reciprocal of the standard deviation of the data, is
A
:
x
↦
2
x
{\displaystyle A:x\mapsto 2{\sqrt {x}}\,}
which, while it is not quite so good at stabilizing the variance, has the advantage of being more easily understood.
Indeed, from the delta method,
V
[
2
x
]
≈
(
d
(
2
m
)
d
m
)
2
V
[
x
]
=
(
1
m
)
2
m
=
1
{\displaystyle V[2{\sqrt {x}}]\approx \left({\frac {d(2{\sqrt {m}})}{dm}}\right)^{2}V[x]=\left({\frac {1}{\sqrt {m}}}\right)^{2}m=1}
.
Generalization
While the Anscombe transform is appropriate for pure Poisson data, in many applications the data presents also an additive Gaussian component. These cases are treated by a Generalized Anscombe transform and its asymptotically unbiased or exact unbiased inverses.
See also
Variance-stabilizing transformation
Box–Cox transformation
References
Further reading
Starck, J.-L.; Murtagh, F. (2001), "Astronomical image and signal processing: looking at noise, information and scale", Signal Processing Magazine, IEEE, vol. 18, no. 2, pp. 30–40, Bibcode:2001ISPM...18...30S, doi:10.1109/79.916319, S2CID 13210703