Digital in-line holography in thick optical systems:

application to visualization in pipes

Nicolas Verrier, Coëtmellec Sébastien, Brunel Marc, Lebrun Denis

To cite this version:

Nicolas Verrier, Coëtmellec Sébastien, Brunel Marc, Lebrun Denis. Digital in-line holography in thick
optical systems: application to visualization in pipes. Applied optics, Optical Society of America,
2008, 47, pp.4147. <10.1364/AO.47.004147>. <hal-00767626>

HAL Id: hal-00767626

https://hal.archives-ouvertes.fr/hal-00767626
Submitted on 20 Dec 2012

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
 Digital in-line holography in thick optical systems:

application to visualization in pipes

N. Verrier, S. Coëtmellec, M. Brunel and D. Lebrun

Groupe d’Optique et d’Optoélectronique, UMR-6614 CORIA, Av. de l’Université,

76801 Saint-Etienne du Rouvray cedex, France

coetmellec@coria.fr, verrier@coria.fr

In this paper we apply digital in-line holography to image opaque objects

through a thick plano-concave pipe. Opaque fibers and opaque particles are

considered. Analytical expression of the intensity distribution in the CCD sensor plane

is derived using generalized Fresnel transform. The proposed model has the ability

to deal with various pipe shape and thickness and compensates for the lack of

versatility of classical DIH models. Holograms obtained with a 12 mm thick

plano-concave pipe are then reconstructed using fractional Fourier transform

(FRFT). This method allows us to get rid of astigmatism. Numerical and experimental

c 2012 Optical Society of America

results are presented.

OCIS codes: 090.0090, 070.0070, 100.0100

1. Introduction

Digital in-line holography (DIH) is a recognized optical technique for flow measurements in trans-

parent liquid media. This method is widely used in various domains such as fluid mechanics [1, 2]

where the flow is seeded with small particles, or in the biological microscopic imaging field [3, 4].

Nevertheless, a transparent pipe can be viewed as an optical system that can introduce

aberrations such as astigmatism [5]. These aberrations make difficult to image objects

1
in the pipe. Compensation for these unwanted effects has been widely investigated.

De Nicola et al. [6] proposed a numerical method to compensate for anamorphism. By

modifying the chirp function and the spatial frequency term of the Fresnel integral

the authors managed to compensate for a severe anamorphism brought by a reflex-

ion diffraction grating. Numerical methods for astigmatism compensation have been

proposed in Refs. [6–8]. Here the authors used a modified chirp function where two

different propagation distances are considered. Astigmatism can also be compensated

for by optical means. For instance, the authors of [9] used index matching to mini-

mize astigmatism: the thin-lens like effect of an ampule filled with contaminants was

investigated to determine index matching parameters for astigmatism compensation.

However, the pipe used to carry a flow must be considered as a thick and cylindrical op-

tical system. Recently, an analytical solution of the scalar diffraction produced by an opaque disk

illuminated by an elliptical, astigmatic and Gaussian beam under Fresnel approximation has been

proposed [10]. In this publication, the astigmatism was introduced and controlled by using a thin

plano-convex cylindrical lens. Using fractional Fourier transformation (FRFT), authors managed

to retrieve a correct image of an opaque disk from astigmatic hologram. It was shown that FRFT

is therefore well suited for these studies where classical methods such as Fresnel transform [11] or

Wavelet transform [12] fail.

The astigmatism introduced by thin lenses is thus well controlled. However, these DIH models can

not be applied to hugger pipes. As a matter of fact, the optical thickness of these has to be taken

into account.

In this paper, the aim is to apply DIH to thick pipe systems. In the first part of the paper, the

intensity distribution, in the CCD plane, of the diffracted field produced by an opaque object located

in a pipe is calculated. A method of Gaussian functions superposition is used to describe

the object function. In the second part, definition of the FRFT is recalled and we demonstrate

2
its ability to reconstruct holograms recorded in such systems. Finally simulations and experiments

are performed to illustrate our results.

2. In-line Holography through pipes

Holography aims to record, on a CCD camera without objective lens, the intensity distribution of the

diffraction pattern of an object illuminated by a monochromatic continuous wave [13]. The numerical

and experimental set-up is represented in Fig. (1). The incident Gaussian beam propagates in free

space over a distance zp and illuminates the pipe. Here, the pipe is modeled as two thick lenses. Their

thickness is denoted by e. The opaque object is located between these two lenses: at a distance

δ from the first thick lens, and at a distance zi from the second lens. The CCD sensor

is located at the distance z next the pipe and records the intensity distribution of the diffraction

pattern.

2.A. Intensity distribution in the CCD sensor plane

In this part, we consider the propagation of a Gaussian beam through our optical system. In the

beam waist plane, located at a distance −zp from the pipe, the Gaussian beam, denoted G, is defined

by:
µ2 + ν 2
 
G (µ, ν) = exp − , (1)
w2

where ω is the waist width and (µ, ν) are the coordinates in the beam waist plane.

The propagation of the Gaussian beam through the pipe, to the CCD sensor is decomposed into

two steps. The first step is the propagation of the illuminating beam (i.e G (µ, ν)) from the beam

waist plane to the front of the opaque object. Using the generalized Huygens-Fresnel integral [14–16],

the amplitude of the field in the pipe, denoted G1 , is defined by:

exp i 2π

Z
λ E1
 
π x 2 x 2


G1 (ξ, η) = G (µ, ν) exp i A1 µ − 2ξµ + D1 ξ
iλ B1x B1y R2 λB1x
p
 
π y 2 y 2

× exp i A ν − 2ην + D η dµdν, (2)
λB1y 1 1

3
Ax,y x,y x,y
1 , B1 , D 1 are given by M1x,y matrices of Eq. (38) in App. (A). To obtain the relation of

the intensity distribution, the linear canonical transformation is used. For the basic definitions and

properties of the linear canonical transformation, we refer to [17]. This transformation is a parame-

terized integral operator of parameters A, B, C and D. Each parameter represents a coefficient of a

transfer matrix denoted M (Cf appendix A). In our case the Fresnel transform is considered [18,19].

It should be noted that the values of the coefficients are different in both ξ and η direction. This is

due to the cylindrical geometry of the pipe. The distance E1 = zp + e + δ corresponds to the waist-

object distance. Using the same formalism, the amplitude in the CCD plane, denoted G2 (x, y), can

be written as:

exp i 2π

Z
E2
 
π x 2 x 2
p λ y


G2 (x, y) = G1 (ξ, η) [1 − T (ξ, η)] exp i A2 ξ − 2xξ + D 2 x
iλ B2x B2 R2 λB2x
 
π y 2 y 2

× exp i A2 η − 2yη + D2 y dξdη, (3)
λB2y

where the parameters Ax,y x,y x,y

2 , B2 , D 2 are given by M2x,y matrices of Eq. (39). The distance E2 =

zi + e + z is the distance between the object and the CCD sensor.

The spatial transmittance of the opaque 2D-object is defined by [1 − T (ξ, η)]. Here T (ξ, η) can

be expressed as a superposition of Gaussian functions [20, 21], such as:

N
X
Ak exp −rT RT Pk Rr ,


T (r) = (4)
k=1

with  
 cos θ sin θ 
R=

,
 (5)
− sin θ cos θ

and  
Bk
 a2 0 
Pk = 

.
 (6)
Bk
0 b2

This expression permits to deal with non symmetrical optical systems and elliptical opaque ob-

jects [22–25]. This Gaussian decomposition is very convenient; as a matter of fact it allows us to

4
establish analytical expression of the diffracted pattern in the CCD plane. It also allows us to

simulate holograms of 3D opaque objects whose 2D projection is elliptic or circular [26]

(e.g. spheroids, fibers ...).

The object parameters are illustrated on Fig. (2). The coefficients a and b are the object radii

within η and ξ axis respectively, θ is the angle between elliptical aperture principal axis and ξ axis.

The Ak and Bk coefficients are determined by numerical resolution of the Kirchhoff equation [21].

Let Rell = b/a representing the particle ellipticity. Considering Rell = 1 lead to the simulation of a

circular particle, whereas Rell 6= 1 is used to simulate elliptical particles. The particular case of the

opaque fiber, parallel to the pipe axis, is obtained when Rell → 0.

In further developments, θ will be considered to be π/2. With this assumption T (ξ, η) becomes:

N  
X Bk
Ak exp − 2 ξ 2 + Rell
2
η2 .


T (ξ, η) = (7)
b
k=1

To simulate our opaque objects, N is fixed to 10. From Eq. (3), G2 (x, y) is split into two integrals

denoted R (x, y) for the reference beam and O (x, y) for the object beam so that :

exp i 2π


λ E2
G2 (x, y) = [R (x, y) − O (x, y)] , (8)
iλ B2x B2y
p

with

 
π
Z
x 2 x 2


R (x, y) = G1 (ξ, η) exp i A2 ξ − 2xξ + D2 x
R2 λB2x
 
π y 2 y 2

× exp i A2 η − 2yη + D2 y dξdη, (9)
λB2x

and

 
π
Z
x 2 x 2


O (x, y) = G1 (ξ, η) T (ξ, η) exp i A2 ξ − 2xξ + D2 x
R2 λB2x
 
π y 2 y 2

× exp i A2 η − 2yη + D2 y dξdη. (10)
λB2y

5
 The functions R (x, y) and O (x, y) are respectively given by:

exp i 2π


λ E1
R (x, y) = K1x K1y K2x K2y
iλ B1x B1y
p
     
π Nx 2 Ny 2 π Mx 2 My 2
× exp − x + yy exp i x + yy , (11)
λ B2x B2 λ B2x B2

and

 N
exp i 2π π D2 2 D2y 2 X

  x
λ E1 x y x y
O (x, y) = p y K1 K1 exp i λ x x + By y Ak K2 eq K2 eq
x
iλ B1 B1 B 2 2 k=1

π Nxeq 2 Nyeq 2 π Mxeq 2 Myeq 2

     
× exp − x + y y exp i x + y . (12)
λ B2x B2 λ B2x B2y

Values of the different parameters of Eqs. (11) and (12) are defined in App. (B).

After theoretical developments the intensity distribution, denoted I (x, y), recorded by the CCD

sensor is:
1 
2  2

I (x, y) = G2 (x, y) G2 (x, y) = y |R| − 2ℜ RO + |O| , (13)
λ2 B2x B2
2
where the upper bar denotes the complex conjugate and ℜ the real part. The square modulus |R|

2
corresponds to the directly transmitted beam whereas |O| is associated with the diffracted part of

the beam.

To illustrate Eq. (13), the optical set-up given on Fig. (1) is considered. We investigate the diffrac-

tion pattern obtained with an opaque fiber parallel to the pipe axis. From the Gaussian decom-

position of the object function, opaque fibers, parallel to the pipe axis, can be obtained taking

Rell → 0. The value of the opaque fiber diameter is 2b = 51.8 m and a → ∞. The glass-made

(n1 = 1.5) pipe is filled with water of refractive index n2 = 1.33. The beam propagates in free space

over zp = 325 mm. The dimensions δ and z are fixed to 18 mm and 23 mm respectively.

The image of Fig. (3) illustrates the intensity distribution of the diffracted field. The simulation

is carried out by calculating the gray level of each pixel using Eq. (13). The size of the hologram is

768×576 pixels, and pixel pitch is 11 m. To validate our simulation, an experiment, which consists in

placing an opaque fiber in the pipe, was performed. The hologram of Fig (4) represents the intensity

6
of the diffraction recorded with the CCD sensor using the same parameters than previously. This

illustration reveals a good accordance between numerical and experimental diffraction patterns. To

confirm this point, the transverse intensity profiles obtained from Fig. (3) and Fig. (4) are presented

on Fig (5). Here the normalized intensity I ∗ is plotted against the x-axis. The intensity profiles are

calculated by cumulating the gray levels along the η-axis (direction of the fiber) over 50 rows, which

are figured out by the rectangular selection on Fig. (4). The diffraction pattern of Fig.

(4) was shifted by 75 m along x-axis so that numerical and experimental results can be

compared. This figure shows the good agreement between numerical and experimental data.

2.B. Case of thin lenses

We have proposed a theoretical model allowing us to deal with propagation of a laser beam through

a thick pipe. In a former study, the effect of cylindrical lenses on the diffraction pattern of a particle

has been investigated in details, leading to the expression of the intensity distribution in the CCD

plane [10]. Using this approach we have a great opportunity to confirm our pipe model results.

In the following developments we aim to compare our thick and cylindrical pipe model with a thin

cylindrical lenses approach. The phase transformation due to a thin lens is [10, 18]:

π x2l y2
  
Φ (xl , yl ) = exp −i + l . (14)
λ fx fy

Here (xl , yl ) are the coordinates in the lens plane and fx,y are the focal lengths of the lens in both

directions. Using Huygens-Fresnel integral [18] the intensity distribution in the sensor plane is:
 2
 exp i 2π E

λ 2
Ithin (x, y) =  [Rthin (x, y) − Othin (x, y)] . (15)
 
 iλE2 

Here, Rthin and Othin can be expressed in a quite similar form than amplitude distributions given

in Eqs. (11) and (12).

The transverse intensity distribution profile presented in Fig. (6) shows:

lim I (x, y) = Ithin (x, y) , (16)

e→0

7
remembering that e represents the glass thickness. This term is contained in matrices

M1x,y and M2x,y (see Eqs. (38) and (39)).

For this example: 2b = 51.8 m, a → ∞, zp = 325 mm, n2 = 1, δ = 18 mm and z = 23 mm. The

comparison with other parameter values leads to the same conclusion. As a result, we can consider

that our results are consistent with those of Ref. [10] and that our model is versatile enough to deal

with various pipe shapes and thickness.

In this section a numerical model allowing to treat thick optical systems has been presented. We

now aim to reconstruct the image of the object by means of FRFT from the calculated intensity

distribution of the diffraction pattern (Eq. (13)). In section 3, mathematical definition of the FRFT

is recalled leading to the reconstruction of holograms recorded in pipe systems.

3. Fractional Fourier transformation analysis of in-line holograms

3.A. Two-dimensional Fractional Fourier transformation

The FRFT is a generalization of the classical Fourier transform. This integral operator has numerous

applications in signal processing [17]. Its mathematical expression is the following [27–29]: the FRFT

of order ax = (2αx )/π and ay = (2αy )/π (for x and y cross section respectively), with 0 ≤ |αx | ≤ π/2

and 0 ≤ |αy | ≤ π/2, of a two dimensional function I (x, y) is

Z
Fαx ,αy [I(x, y)](xa , ya ) = Nαx (x, xa ) Nαy (y, ya )I(x, y) dx dy, (17)
R2

where the kernel of the fractional operator is defined by

x2 + x2a
   
i2πxa x
Nαp (x, xa ) = C(αp ) exp iπ 2 exp − 2 , (18)
sp tan αp sp sin αp

and
αp
exp(−i( π4 sign(sin αp ) − 2 ))
C(αp ) = . (19)
|s2p sin αp |1/2

Here p = x, y. Generally, the parameter sp is considered as a normalization constant. In our case,

its value is defined from the experimental set-up according to [30]: s2p = Nppx · δp2 . The number

8
of samples is Nppx in both intensity distribution I (x, y) and fractional domain. δp is the sampling

period along the two axes of the image. C(αp ), which is a function of the fractional order, insures the

energy conservation law to be valid. Discretization of the FRFT kernel is performed using

an orthogonal projection method proposed by Pei et al. [31].

3.B. Optimal fractional orders to refocus numerically over the object

In our case, the numerical reconstruction can be considered as a numerical refocusing over the


object [32]. To do that, the quadratic phase, denoted ϕ = arg RO , and contained in the term

2ℜ RO of Eq. (13) must be evaluated. This term is composed of a linear chirp modulated by a

sum of complex Gaussian functions. Information about the distance between the CCD sensor and the

object is carried by the linear chirp (B2x,y ) whereas information about object size is carried by the

modulation (Mx,yeq and Nx,yeq ). The optimal reconstruction of the image consists in compensating

the quadratic phase terms of the intensity distribution [10].

The quadratic phase can be determined from Eqs. (11) and (12)

Mx − D2x My − D2y
    
π
ϕ= x2 + y2 , (20)
λ B2x B2y


thus ℜ RO can be written as:
  
ℜ RO = RO cos (iϕ) . (21)

The quadratic phase term contained in the FRFT kernel, denoted ϕa , is given by

 
cot αx 2 cot αy 2
ϕa = π x + y . (22)
s2x s2y

The image reconstruction is obtained by applying the FRFT to the intensity distribution of Eq.

(13):

2 2  
Fαx ,αy [I(x, y)] ∝ Fαx ,αy [|R| + |O| ] − 2Fαx ,αy [RO cos ϕ]. (23)

2 2
The terms |R| and |O| contain no linear chirps, thus they do not influence the optimal fractional

orders of reconstruction to be determined. Only the second term is useful for image reconstruction.

9
By noting that 2 cos ϕ = exp (−iϕ) + exp (iϕ), Eq. (23) becomes:

2 2
Fαx ,αy [I(x, y)] ∝ Fαx ,αy [|R| + |O| ]
  
xa x ya y
Z
 
− C (αx ) C (αy ) RO exp [i (ϕa − ϕ)] exp −i2π + 2 dxdy
R2 s2x sin αx sy sin αy
  
xa x ya y
Z
 
− C (αx ) C (αy ) RO exp [i (ϕa + ϕ)] exp −i2π + dxdy. (24)
R2 s2x sin αx s2y sin αy

The best hologram reconstruction is reached when one of the quadratic phase term is brought to

zero. Thus reconstruction is performed if

ϕa ± ϕ = 0. (25)

The optimal fractional orders αopt

x and αopt
y are defined from Eqs. (20), (22) and (25) and take the

values:
B2x λ B2y λ
   
αopt = arctan ∓ , αopt = arctan ∓ . (26)
x
s2x (Mx − D2x ) y
s2y (My − D2y )

It should be noted that we have checked that if we consider M2x,y to be free space propagation,

leading to B2x,y = zi + e + z and D2x,y = 1, the expression of the optimal fractional order of Ref. [10]

are recovered.

3.C. Numerical simulations

Experimental set-up of Fig. (1) is used to perform simulations. An application of the model is

illustrated on Fig. (3). Recall that we consider an opaque fiber (Rell → 0) 2b = 51.8µm in width.

The glass made pipe is filled with water (n2 = 1.33). δ and z are set to 18 mm and 23 mm

respectively. Rx1 , Ry1 , Ry2 → ∞ and Rx2 = 18 mm are the first thick lens curvatures along x- and

y-axis. The curvature radii of the second thick lens have the same values but opposite signs. With

these parameters and owing to Eq. (26) the optimal fractional orders within this configuration are

aopt
x = 0.337, aopt
y = 0.273. (27)

The reconstruction of the fiber with FRFT can be seen on Fig. (7). As we can see on this figure, the

reconstructed image is disturbed by background fringes. This effect is due to the in-line configuration

10
and is commonly known as the twin image effect [33].

3.D. Experimental results

Since now, we have presented a method to simulate holograms in thick optical systems. Reconstruc-

tion has been successfully performed thanks to FRFT. In order to validate theoretical developments

and simulations, a glass pipe with curvatures Rx1 , Ry1 , Ry2 → ∞ and Rx2 = 18 mm has been used

for the experiments. Curvatures on the other side of the pipe are deduced from Rx1 , Ry1 , Rx2 , Ry2

by taking negative values.

The image of Fig. (4) represents the intensity of the diffraction pattern recorded with a 768 ×

576 px2 CCD sensor with 11 µm pitch. The object is a 2b = 51.8µm opaque fiber, δ and z are

approximately and respectively equal to 18 mm and 23 mm. Theoretical orders associated with this

experiment are given by Eq. (27). To perform FRFT on this experimental image, FRFT orders have

been adjusted to obtain the best image of the fiber i. e the best contrast between the reconstructed

fiber and the background. Thus doing, the accuracy on the fractional order is approximately 10−2 .

Reconstruction is presented on Fig. (8), estimated optimal fractional orders are ax = 0.33 and

ay = 0.27. These values are very close to the theoretical ones.

This good accordance is confirmed by the curves of Fig. (9). Here theoretical values of aopt
x (solid

line) and aopt

y (dashed line) are plotted versus the distance between the pipe and the CCD sensor.

Theoretical values corresponds to an air filled glass pipe, δ is fixed to 10 mm and z varies from 0

to 140 mm. Note that, when z = 0, ax,y 6= 0, this is due to the fact that, in this case, z represents

the distance between the pipe and the CCD sensor instead of the distance between the object and

the CCD sensor which is considered in [10]. Experiments performed with this configuration allow us

to plot estimated values of aopt opt

x (circles) and ay (diamonds) on the same graph. As shown by Fig.

(9), estimated and theoretical values are closely linked. Theoretical representation hereby presented

is therefore well adapted to these optical systems.

11
 Until then, we have performed reconstruction with objects for which position is well

known. We now apply our formalism to reconstruct hologram of latex beads. The

intensity distribution of the diffracted field is represented on Fig. (10). Here, 100 m

in diameter latex beads are used, zp = 325 mm (i.e. propagation distance between the

source and the pipe) and z = 23 mm (i.e. distance between the pipe and the CCD

sensor). This study gives us the opportunity to compare reconstruction using our novel

approach with reconstruction using a thin lens approach. As far as every parameter

of the experimental set-up is known, excepting δ, the reconstruction process is the

following: for δ ∈ [0, 36 mm] we calculate the optimal FRFT orders using Eq. (26). Doing

so, we are able to refocus on the particles which are in the pipe.

In Fig. (6), it is shown that when the pipe thickness e is close to zero, the two

approaches are equivalent. Thus, to reconstruct the hologram of Fig. (10) with the thin

lens approach, we only need to use the orders given in Eq. (26) with e = 0.

Comparison between the two reconstruction methods can be made with Fig. (11).

Here, the reconstruction is realized with the same value of δ in both cases. We can notice

that refocusing on the particle is impossible with the thin lens approach, whereas our

thick lens approach allows a good reconstruction of the image of particles.

Therefore, the thick lens formalism is well adapted to pipe flow studies: as we are

able to refocus on particles in a pipe, metrologies of particles diameter is possible.

4. Conclusion

In this paper, an analytical expression of the scalar diffraction, under Fresnel approximations, in

thick optical systems such as pipes is derived. As such optical systems reveals astigmatism, FRFT is

a reliable tool to perform hologram reconstruction. Optimal fractional orders have been calculated

leading to a satisfactory reconstruction of either numerical or experimental images. Numerical ex-

12
periments have been performed showing a good agreement with experimental results. Comparison

between thin lens and thick lens approaches has been performed showing that the thin

lens formalism is not well adapted to the study of pipe flows.

Acknowledgments

These works come within the scope of Hydro-Testing Alliance (HTA) with the support of the Euro-

pean Commission’s Sixth Framework Programme under DG Research.

13
A. Appendix A : Transfer matrices of the optical system

Each part of an optical system can be represented by a matrix. Using this principle, we can build a

set of matrices corresponding to our experiments (Fig. (1)). After the beam shaping step, the beam

propagates in free space over a distance zp . The associated Mzp matrix is:
 
 1 zp 
M zp = 

.
 (28)
0 1

Then the beam is refracted at the interface between free space and glass (n1 = 1.5). The curvatures

along both directions are Rx1 and Ry1 . MRx1 , MRy1 matrices along x and y axis are:
   
 1 0   1 0 
MRx1 = 

,
 MR1y = 

.
 (29)
n0 −n1 n0 −n1
Rx1 1 Ry1 1

After refraction, the beam propagates in glass (n1 =1.5). For this step:
 
e
 1 n1 
Me = 

,
 (30)
0 1

Next step is the refraction of the beam at the interface between glass and the medium inside the

pipe (refractive index n2 ). Curvature are given by Rx2 and Ry2 in x and y direction.
   
 1 0   1 0 
MRx2 = 

,
 MRy2 = 

,
 (31)
n1 −n2 n1 −n2
Rx2 1 Ry2 1

We are now in the pipe. To reach the object, we have to propagate over δ.
 
δ
 1 n2 
Mδ = 

.
 (32)
0 1

Doing the same over zi permit to reach the output of the pipe:
 
zi
 1 n2 
M zi =

.
 (33)
0 1

14
 After a refraction at the interface (curvature Rx3 and Ry3 ) between the medium with refractive

index n2 and glass:

   
 1 0   1 0 
MRx3 = 

,
 MRy3 = 

,
 (34)
n2 −n1 n2 −n1
Rx3 1 Ry3 1

a propagation in glass (n1 =1.5)

 
e
 1 n1 
Me = 

,
 (35)
0 1

a refraction at the interface (curvature Rx4 and Ry4 ) between glass (n1 =1.5) and free space:
   
 1 0   1 0 
MRx4 = 

,
 MRy4 = 

,
 (36)
n1 −n0 n1 −n0
Rx4 1 Ry4 1

and free space propagation over z:

 
 1 z 
Mz = 

,
 (37)
0 1

the whole ABCD system is described.

B. Appendix B : Amplitude distributions R (x, y) and O (x, y)

In Appendix A each part of our optical system has been represented with transfer matrices. Using

this formalism allows us, under paraxial conditions, to deal with propagation of a Gaussian point

source through the pipe.

Intensity distribution of the diffraction pattern in the CCD sensor plane is determined by consid-

ering two matrix systems: M1x,y and M2x,y .

M1x,y is composed of three steps : propagation in free space over zp , propagation through the first

thick lens, propagation in a medium of refractive index n2 over δ. It is characterized by two transfer

matrices:  
Ax,y
1 B1x,y
M1x,y = Mδ × MLx,y
 
1
× M zp = 

.
 (38)
C1x,y D1x,y

15
 M2x,y is also composed of three steps : propagation in a medium of refractive index n2 over zi ,

propagation through the second thick lens, propagation in free space over δ. Transfer matrices for

this system are:

 
x,y
 A2 B2x,y 
M2x,y = Mz × MLx,y
2
× M zi =
 .
 (39)
C2x,y D2x,y

Thanks to Eqs. (29), (30), (31) one can build the transfer matrices of the first thick lens:

MLx1 = MRx2 × Me × MRx1 , MLy 1 = MRy2 × Me × MRy1 . (40)

By using the same method for the second lens, we obtain:

MLx2 = MRx4 × Me × MRx3 , MLy 2 = MRy4 × Me × MRy3 . (41)

B.A. Propagation through M1x,y

After analytical developments of Eq. (2), the complex amplitude distribution in the object plane is:
" !#
exp i 2π


E ξ2 η2 ξ2 η2
  
1 y π
G1 (ξ, η) = p λ K1x K1 exp − + 2 exp −i + , (42)
iλ B1x B1y 2
ω1x ω1y λ R1x R1y

with K1x,y given by:

!1/2
πω 2
K1x,y = , (43)
1 − iAx,y πω 2
1 λB x,y 1

ω1x,y and R1x,y are respectively the beam radii and the wavefront curvature in the particle plane.

Their mathematical expressions are:

" 2 #1/2
λB1x,y πω 2 B1x,y
 
ω1x,y = 1+ Ax,y , R1x,y = − . (44)
1
λB1x,y
2
πω

πω2
Ax,y
1 x,y
D1x,y −
λB1
 2
πω2
1+ Ax,y
1 x,y
λB
1

B.B. Propagation through M2x,y

Propagation to the CCD sensor plane is calculated thanks to the generalized Huygens-Fresnel integral

(see Eq. (3)).

16
B.B.1. Amplitude distribution R (x, y)

R is associated with the reference wave. After analytical developments of Eq. (9), the amplitude

distribution R is:

exp i 2π


λ E1
R (x, y) = K1x K1y K2x K2y
iλ B1x B1y
p
     
π Nx 2 Ny 2 π Mx 2 My 2
× exp − x + y exp i x + y , (45)
λ B2x B2y λ B2x B2y

where,
 2 
πω12x,y B2x,y

λB2x,y R1x,y − Ax,y
2
πω12x,y
λB2x,y
Mx,y = D2x,y +  2  2 , Nx,y =  2  2 , (46)
πω12x,y B2x,y πω12x,y B2x,y
1+ λB2x,y R1x,y − Ax,y
2 1+ λB2x,y R1x,y − Ax,y
2

and  1/2
πω12x,y
K2x,y =  . (47)
 
πω12x,y
 x,y 
B2
1 + i λB x,y R1x,y − Ax,y
2
2

B.B.2. Amplitude distribution O (x, y)

O is the amplitude of the diffracted wave. We define ω1x,y eq and R1x,y eq

1 1 ℜ{Bk } 1 1 2 ℜ{Bk }
= + , = + Rell , (48)
ω12x eq 2
ω1x b2 ω12y 2
ω1y b2
eq

and
1 1 ℑ{Bk }λ 1 1 2 ℑ{Bk }λ
= + , = + Rell , (49)
R1x eq R1x πb2 R1y eq R1y πb2

to simplify notations. It should be noted that ℜ and ℑ stand for real and imaginary part respectively.

Thus O becomes:

 N
exp i 2π π D2 2 D2y 2 X

  x
λ E1 x y x y
O (x, y) = K K exp i x + y Ak K2 eq K2 eq
iλ B1x B1y
1 1
λ B2x B2y
p
k=1

π Nxeq 2 Nyeq 2 π Mxeq 2 Myeq 2

     
× exp − x + y exp i x + y , (50)
λ B2x B2y λ B2x B2y

17
with
2 
πω12x,y
 
B2x,y
eq
λB2x,y R1x,y eq − Ax,y
2
Mx,yeq =  2  2 ,
πω12x,y B2x,y
1+ eq
λB2x,y R1x,y eq − Ax,y
2

πω12x,y
eq
λB2x,y
Nx,yeq =  2  2 , (51)
πω12x,y B2x,y
1+ eq
λB2x,y R1x,y eq − Ax,y
2

and  1/2
x,yeq  πω12x,yeq 
K2 = πω12x,y
  . (52)
B2x,y
− Ax,y
 
1+i λB2x,y
eq
R1x,yeq 2

18
References

1. G. Pan, H. Meng, ”Digital holography of particle fields: reconstruction by use of complex am-

plitude,” Appl. Opt. 42, 827-833 (2003).

2. M. Malek, D. Allano, S. Cotmellec, C. zkul, D. Lebrun, ”Digital in-line holography for three-

dimensional-two-components particle tracking velocimetry,” Meas. Sci. Technol. 15, 699-705

(2004).

3. E. Malkiel, J. Sheng, J. Katz, J. Strickler, ”The three-dimensional flow field generated by a

feeding calanoid copepod measured using digital holography,” J. Exp. Bio. 206, 3657-3666

(2003).

4. J. Garcia-Sucerquia, W. Xu, S. Jericho, P. Klages, M. Jericho, H. Kreuzer, ”Digital in-line

holographic microscopy,” Appl. Opt. 45, 836-850 (2006).

5. C.S. Vikram, ”Particle field holography,” Cambridge Studies in Modern Optics, (1992).

6. S. De Nicola, P. Ferraro, A. Finizio, G. Pierattini, ”Correct-image reconstruction in the presence

of severe anamorphism by means of digital holography,” Opt. Lett. 26, 974-976 (2001).

7. S. Grilli, P. Ferraro, S. De Nicola, A. Finizio, G. Pierattini, R. Meucci, ”Whole optical wavefields

reconstruction by Digital Holography,” Opt. Express 9, 294-302 (2001).

8. S. De Nicola, P. Ferraro, A. Finizio, G. Pierattini, ”Wave front reconstruction of Fresnel off-axis

holograms with compensation of aberrations by means of phase-shifting digital holography,”

Opt. Laser Eng. 37, 331-340 (2002).

9. J.S. Crane, P. Dunn, B.J. Thompson, J.Z. Knapp, J. Zeiss, ”Far-field holography of ampule

contaminants,” Appl. Opt. 21, 2548-2553 (1982).

10. F. Nicolas, S. Cotmellec, M. Brunel, D. Allano, D. Lebrun, and A. J. Janssen, ”Application of

the fractional Fourier transformation to digital holography recorded by an elliptical, astigmatic

Gaussian beam,” J. Opt. Soc. Am. A 22, 2569-2577 (2005).

19
11. L. Onural, P. Scott, ”Digital decoding of in-line holograms,” Opt. Eng. 26, 1124-1132 (1987).

12. L. Onural, ”Diffraction from a wavelet point of view,” Opt. Lett. 18, 846-848 (1993).

13. F. Nicolas, S. Cotmellec, M. Brunel and D. Lebrun, ”Digital in-line holography with a sub-

picosecond laser beam,” Optics Communications 268, 27-33 (2006).

14. C. Palma, V. Bagini, ”Extension of the Fresnel transform to ABCD systems,” J. Opt. Soc. Am.

A 14, 1774-1779 (1997).

15. A.J. Lambert, D. Fraser, ”Linear systems approach to simulation of optical diffraction,” Appl.

Opt. 37, 7933-7939 (1998).

16. H.T. Yura, S.G. Hanson, ”Optical beam wave propagation through complex optical systems,”

J. Opt. Soc. Am. A 4, 1931-1948 (1987).

17. H.M. Ozaktas, Z. Zalevsky, M.A. Kutay, The Fractional Fourier Transform: with Applications

in Optics and Signal Processing, Wiley, (2001).

18. J.W. Goodman, Introduction to Fourier Optics, Roberts and Company, third edition, (2005).

19. A.E. Siegman, Lasers, University Science Books, Sausalito, California, (1986).

20. J.J. Wen, M. Breazeale, ”Gaussian beam functions as a base function set for acoustical field

calculations,” Proc IEEE Ultrasonics Symposium, 1137-1140 (1987).

21. J.J. Wen, M. Breazeale, ”A diffraction beam expressed as the superposition of Gaussian beams,”

J. Acoust. Soc. Am. 83, 1752-1756 (1988).

22. C. Zheng, D. Zhao, X. Du, ”Analytical expression of elliptical Gaussian beams through non-

symmetric systems with an elliptical aperture,” Optik 117, 296-298 (2005).

23. X. Du, D. Zhao, ”Propagation of decentered elliptical Gaussian beams in apertured and non-

symmetrical optical systems,” J. Opt. Soc. Am. A 23, 625-631 (2006).

24. X. Du, D. Zhao, ”Propagation of elliptical Gaussian beams in apertured and misaligned optical

systems,” J. Opt. Soc. Am. A 23, 1946-1950 (2006).

20
25. X. Du, D. Zhao, ”Propagation of elliptical Gaussian beams modulated by an elliptical annular

aperture,” J. Opt. Soc. Am. A 24, 444-450 (2007).

26. F. Slimani, G. Grehan, G. Gouesbet, D. Allano, ”Near-field Lorenz-Mie theory and its applica-

tion to microholography,” Appl. Opt. 23, 4140-4148 (1984).

27. A.C. McBride, F.H. Kerr, ”On Namias’s fractional Fourier transforms,” IMA J. Appl. Math 39,

159-175 (1987).

28. V. Namias, ”The fractional order Fourier transform and its application to quantum mechanics,”

J. Inst. Maths Its Applics 25, 241-265 (1980).

29. A.W. Lohmann, ”Image rotation, Wigner rotation, and the fractional Fourier transform,” J.

Opt. Soc. Am. A 10, 2180-2186 (1993).

30. D. Mas, J. Pérez, C. Hernández, C. Vázquez, J. J. Miret and C. Illueca, ”Fast numerical calcu-

lation of Fresnel patterns in convergent systems,” Optics Communications 227, 245-258 (2003).

31. S-C Pei, M-H Yeh, C-C Tseng, ”Discrete Fractional Fourier Transform Based on Orthogonal

Projections,” IEEE Trans. Signal Processing 47, 1335-1348 (1999).

32. N. Verrier, S. Coëtmellec, M. Brunel, D. Lebrun and A.J.E.M. Janssen, ”Digital in-line hologra-

phy with an elliptical, astigmatic Gaussian beam: wide-angle reconstruction,” J. Opt. Soc. Am.

A, 25, 1459-1466 (2008).

33. D. Gabor, ”Microscopy by reconstructed wave-fronts,” Proc. Roy. Soc. A. 197, 454-487 (1949)

21
List of Figures

1 a) Schematic representation of the optical set-up (not to scale). Definition of the

numerical and experimental parameters. b) Close-up of the pipe used in simulations

and experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2 Schematic representation of the object. . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3 Simulation of the diffraction pattern of a 51.8 µm opaque fiber, parallel to the axis of

a glass pipe, n2 = 1.33, λ = 632.8nm, z = 23mm, δ = 18mm. . . . . . . . . . . . . . 25

4 Experimental diffraction pattern of a 51.8 µm opaque fiber, parallel to the axis of a

glass pipe, n2 = 1.33, λ = 632.8nm, z = 23mm, δ = 18mm. . . . . . . . . . . . . . . 26

5 Comparison between simulated and experimental intensity distributions . . . . . . . 27

6 Normalized intensity distribution recorded by the CCD: comparison between thin

lens and thick lens models, with e = 0 mm. Simulation parameters are : 2b = 51.8 m,

a → ∞, zp = 325 mm, n2 = 1, δ = 18 mm and z = 23 mm . . . . . . . . . . . . . . 28

7 Reconstruction of the fiber image from the diffraction pattern of Fig. (3) by FRFT

with aopt opt

x = 0.337 and ay = 0.273. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

8 Reconstruction of the fiber image from the diffraction pattern of Fig. (4) by FRFT

with aopt opt

x = 0.33 and ay = 0.27. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

9 Comparison between theoretical fractional orders and optimal fractional orders esti-

mated from the experimental holograms. . . . . . . . . . . . . . . . . . . . . . . . . . 31

10 Experimental diffraction pattern of 100m latex beads. zp = 325 mm, z = 23 mm. . . 32

11 Reconstruction of the latex beads image. (a) Reconstruction with thin lens parame-

ters. (b) Reconstruction using thick lens approach. . . . . . . . . . . . . . . . . . . . 33

22
 (a)

(b)

Fig. 1. a) Schematic representation of the optical set-up (not to scale). Definition of the

numerical and experimental parameters. b) Close-up of the pipe used in simulations and

experiments

23
Fig. 2. Schematic representation of the object.

24
Fig. 3. Simulation of the diffraction pattern of a 51.8 µm opaque fiber, parallel to the axis

of a glass pipe, n2 = 1.33, λ = 632.8nm, z = 23mm, δ = 18mm.

25
Fig. 4. Experimental diffraction pattern of a 51.8 µm opaque fiber, parallel to the axis of a

glass pipe, n2 = 1.33, λ = 632.8nm, z = 23mm, δ = 18mm.

26
 1
Theoretical
0.9 Experimental

0.8

0.7

0.6
I*

0.5

0.4

0.3

0.2

0.1

0
−1 −0.5 0 0.5 1
x (m) x 10
−3

Fig. 5. Comparison between simulated and experimental intensity distributions

27
 1
Thick
0.9 Thin

0.8

0.7

0.6
I*

0.5

0.4

0.3

0.2

0.1

0
−1 −0.5 0 0.5 1
x (m) x 10
−3

Fig. 6. Normalized intensity distribution recorded by the CCD: comparison between thin

lens and thick lens models, with e = 0 mm. Simulation parameters are : 2b = 51.8 m,

a → ∞, zp = 325 mm, n2 = 1, δ = 18 mm and z = 23 mm

28
Fig. 7. Reconstruction of the fiber image from the diffraction pattern of Fig. (3) by FRFT

with aopt opt

x = 0.337 and ay = 0.273.

29
Fig. 8. Reconstruction of the fiber image from the diffraction pattern of Fig. (4) by FRFT

with aopt opt

x = 0.33 and ay = 0.27.

30
 1

0.9

0.8

0.7
ax, ay

0.6

0.5

0.4 a th
x
a est
x
0.3 a th
y
a est
y
0.2
0 20 40 60 80 100 120 140

z (mm)
Fig. 9. Comparison between theoretical fractional orders and optimal fractional orders esti-

mated from the experimental holograms.

31
Fig. 10. Experimental diffraction pattern of 100m latex beads. zp = 325 mm, z = 23 mm.

32
 (a) (b)

Fig. 11. Reconstruction of the latex beads image. (a) Reconstruction with thin lens param-

eters. (b) Reconstruction using thick lens approach.

33