May 1, 2014 / Vol. 39, No.

9 / OPTICS LETTERS

2569

Automatic method for focusing biological specimens

in digital lensless holographic microscopy
Carlos A. Trujillo and Jorge Garcia-Sucerquia*
Universidad Nacional de Colombia, Sede Medellin, School of Physics, A.A. 3840 Medellin 050034, Colombia
*Corresponding author: jigarcia@unal.edu.co
Received February 21, 2014; revised March 24, 2014; accepted March 25, 2014;
posted March 26, 2014 (Doc. ID 206833); published April 18, 2014
A self-focusing method applicable to digital lensless holographic microscopy is presented. The method searches for
the global minimum of the area enclosing a given amount of energy in a region surrounding the object of interest.
The proposed modified enclosed energy method has been tested on self-focusing experimental holograms of a paramecium specimen and a section of the head of a drosophila melanogaster fly. The presented self-focusing technique
also has been contrasted with some of the already reported methods to seek the best focus image. 2014 Optical
Society of America
OCIS codes: (090.1995) Digital holography; (260.5950) Self-focusing; (180.6900) Three-dimensional microscopy.
http://dx.doi.org/10.1364/OL.39.002569

Focusing in microscopy plays a key role in the quality of

the studies that use this imaging tool. The first approach
to seek for the best focal plane relies upon the expertise
of the user. The need for more quantitative studies
derived on the development of automated self-focusing
strategies that were applied initially to optical microscopy [1,2]. Some of these strategies were extended to
the scenario of digital holographic microscopy (DHM)
where the numerical recovering of the recorded image
eases the application of self-focusing approaches. A large
variety of methods has been applied or shows potential
to have quantitative focusing of the reconstructed images
[37] including phase difference imaging [8]; most of
these techniques have been applied mainly in off-axis
DHM. For the particular case of digital lensless holographic microscopy (DLHM), which is an in-line architecture, a self-focusing method based on the analysis of the
evolution of the intensity along the direction of propagation has been also developed [9]. This technique, which
has been successfully applied to automatically track
micrometer-sized particles, does not work when the
specimens of interest have complex internal structures
and dimensions from tens to hundreds of micrometers.
The two last characteristics are common to most of
the biological specimens. This means that the current
state of the art of DLHM lacks of an automatic method
to self-focus holograms of objects with these of characteristics. In this Letter, a method based on the law of the
conservation of energy and on the consideration of
the spherical illumination for the automatic focusing of
digital holograms of biological specimens recorded via
DLHM is presented.
In DLHM the sample is illuminated by a spherical
wavefront with a wavelength . A digital screen records
the amplitude superposition of the portion of the spherical wavefront that is scattered by the specimen, U scat ,
with the portion of that wavefront that travels from
the point source to the screen with no distortion, U ref .
The sample (center of the digital screen) is placed at a
distance Z
L from the point source. The intensity recorded by the digital screen is transferred to a PC for
its further processing, which includes the retrieving of
U scat . A schematic DLHM setup is illustrated in Fig. 1.
0146-9592/14/092569-04$15.00/0

In this setup, the point source is produced by focusing

down the light from a laser onto the surface of a pinhole
with diameter close to . In this setup, the recorded
intensity by the digital screen is given by
I
r  jU scat
rj2  jU ref
rj2
 U scat
rU ref
r  U scat
rU ref
r:

(1)

To DLHM properly work, the preferred samples are

weak scatterers [10]. This condition leads to neglect
jU scat
rj2 in Eq. (1). The terms in the square brackets
are the twin images [11]; the utilization of spherical
illumination guarantees that these twin images do not
perturb to each other in the reconstructed plane, as it
has been extensively reported elsewhere [11,12]. The
position vector r 
x; y; L represents a point at the
digital screen.
The intensity recorded by the digital screen with no
sample present is pixel-wise subtracted from Eq. (1) to
~
produce the new intensity I
r.
The scattered wave by
the sample U scat is recovered by computing numerically
the diffraction that a spherical wavefront undergoes as
~
it illuminates I
r.
The very details of this numerical calculation can be read in [13]. The reconstructed amplitude

Fig. 1.

Schematic representation of a DLHM setup.

2014 Optical Society of America

2570

OPTICS LETTERS / Vol. 39, No. 9 / May 1, 2014

U scat
r0  is computed over a plane with points denoted by

the position vector r0 
x0 ; y0 ; Zr. As that diffraction
process is numerically calculated from a hologram with
M M pixels, it reads
U scat
sx0 ; ty0 ; Zr
 x00 y00 expik
s2 x00 x0  t2 y00 y0 2L

M21
X
M21
X

I~ 00
mx00 ; ny00 

mM2 nM2

expik
m2 x00 x0  n2 y00 y0 2L

expik

s m2 x00 x0 
t n2 y00 y0 2L: (2)
In Eq. (2), I~ 00
mx00 ; ny00  is an interpolated intensity
that accounts for the reconstruction distance Zr; m, n,
s, and t are integer numbers to denote the discrete coordinates in the interpolated and reconstruction planes.
This equation represents a scalable convolution that
allows the change of the size of the reconstructed image
to suit the appropriate variation of that size as the spherical wavefront propagates from the point source. In other
words, Eq. (2) allows the setting of the proper scaling
factors introduced by the use of the spherical illuminations, as has been shown elsewhere [11,13].
In DLHM two particles, which are separated by distance r and can be distinguished if r 2NA [11],
with NA the numerical aperture of the microscope. Since
DLHM is a lensless technique, the NA is controlled by the
width W p
of


























the digital screen and the distance L, which is
NA  1 1  4L2 W 2 . To obtain spatial resolution in
the micrometer range in DLHM using visible light, the
width of the digital screen W , using either a CCD or
CMOS camera, should be in the order of its distance
to the point source L. For depth resolution in DLHM,
two points can be resolved in-depth if z 2NA2 [11].
The illumination of the sample by a spherical wavefront imposes that (1) it has to be possible to change numerically the size of the reconstructed plane for the
different distances from the point source, while the number of pixels is fixed and (2) the irradiance of the reconstructed images changes as a function of the distance
from the point source following the well-known 1r 2
law [14]. These two features have to be considered in
the design of an automatic strategy for focusing holograms recorded with DLHM. The former is already
satisfied when the holograms are reconstructed via
Eq. (2). For the latter, the self-focusing method has to
consider that the irradiance between two planes located
at distances Zr i and Zr f from the point source varies as
jU scat
sx0 ; ty0 ; Zr f j2 

Zr 2i
jU scat
sx0 ; ty0 ; Zr i j2 :
Zr 2f
(3)

The condition expressed by Eq. (3) represents the law

of conservation of energy in DLHM, accounting that the
measurable quantity is the irradiance and that the sample
is illuminated by a spherical wavefront. These considerations allow the proposal of a novel method to self-focus

holograms of biological specimens as they are imaged via

DLHM. To distinguish the present self-focusing method
from others reported in the literature, which consider
the law of conservation of energy, we call our technique
modified enclosed energy.
The present method for self-focusing DLHM hologram
searches for the global minimum of the area enclosing a
given amount of energy in a region surrounding the object of interest. The process to achieve the self-focused
image is performed in the following stages:
(1) The user launches the self-focusing technique by
choosing the range of distances Zr i ; Zr f  where the
specimen is supposed to be focused and by selecting
the offset Zr between consecutive reconstructed
images.
(2) The hologram is reconstructed for the distance
Zr i .
(3) Over the reconstructed hologram in (2), the user
chooses a square of interest that encloses the specimen
to be focused.
(4) The algorithm relocates the square with its center
placed at the center of mass of the enclosed irradiance.
(5) The algorithm computes the total enclosed irradiance Ir i within the square of N N pixels selected by the
user
Ir i 

N X
N
X
s1 t1

jU scat
sx0 ; ty0 ; Zr i j2 :

(4)

(6) The algorithm reconstructs the modified intensity

~
I
r
for a number of
Zr f Zr i Zr planes.
(7) For each reconstructed hologram placed at distance Zr n , the algorithm seeks for the square, with its
center at the center of mass of the irradiance, which
encloses the quantity Ir n  Ir i Zr 2i Zr 2n , namely, it
searches for the index P n that makes
n X
n
X
Zr 2i

jU scat
sx0 ; ty0 ; Zr n j2 :
Zr 2n
s1 t1

Ir i

(5)

(8) From a vector formed by the couples fZr n ; P n g, the

algorithm searches for the smallest P n ; the corresponding Zr n is the distance of the best focal plane.
(9) Once the focal plane has been found in stage (8),
the user can do a finer focusing by selecting a narrower
range of distances Zr 0i ; Zr 0f  and a smaller offset Zr 0 .
Within this new range, the stages from (2) to (8) are
repeated to find a finer focused image.
The smallest valid offset should be greater than the
depth resolution z  2NA2 of the microscope, and
it must guarantee that between consecutive images the
sought energy is enclosed within squares that differ in
at least one pixel in side. For DLHM, the latter condition
is achieved if


W Z W Z0

(6)
;
Z
Z0
L
where W Z 0 is the side of the square at the focus plane Z 0 ;
W Z is the side of the square at the reconstruction distance Z, and is the pixel size of the digital screen placed
at a distance L from the point source.

May 1, 2014 / Vol. 39, No. 9 / OPTICS LETTERS

To illustrate the method on self-focusing holograms

of biological specimens, we have utilized it in the
reconstruction of a DLHM hologram of a paramecium.
The hologram was recorded on a CMOS camera with
1024 1024 square pixels, 6.7 m in side. A point source,
emitting at a wavelength of 532 nm, was placed 20 mm
apart from the center of the CMOS camera. For this
experiment, it was observed that the paramecium was
in focus in the range from 1 to 2 mm. The first
reconstruction was done for Zr i  1 mm, and a step
of 0.125 mm was chosen. For that reconstruction, a
square with 417 pixels in side was selected by the user
such that the out-of-focus specimen was completely
enclosed. For that square it was computed that irradiance Ir i used for computing the indices P n for each
Zr n , according Eq. (5). The different frames reconstructed by the algorithm to seek for the best focal planes
are shown in Fig. 2. In each individual frame, the green
square illustrates the area within which the irradiance is
evaluated. The smallest yellow circle corresponds to the
sought specimen; the largest circle is a two-time magnification of the smallest. The side of the square that encloses the amount of energy initially bounded within the
square with 417 pixels in side, starts to shrink steady. As
the reconstruction distance Zr  1.5 mm, that side
reaches the minimum value 371 pixels. Since Zr to
Zr  2.0 mm, the side of the square grows again, indicating that the best focal plane is located for Zr  1.5 mm
as the step is set at 0.125 mm. The reader can directly
compare the images in the large yellow circles to verify
that the image at that reconstruction distance is better
focused than the others. At this stage, the user could
decide to select a narrower range than 1 mm centered
around 1.5 mm to look for an even better-focused image.
For this experiment, the offset could be as small as
10 m; however, for illustration purposes, we limit the
presentation for the case just presented.

Fig. 2. Example of the algorithm operation for the modified

enclosed energy. See text for further details.

2571

The proposed method for self-focusing DLHM holograms has been contrasted with other methods reported
in the literature: Dubois energy method [3], edgedetection techniques as Laplacian, gradient, and variance
[15], spectral analysis of the reconstructed image [15],
and the study of the contrast of the resulting image via
Tamuras coefficient [6] are the metrics that performance
was contrasted with in regards to the modified enclosed
energy presented in this Letter. For this comparison, we
utilized a DLHM hologram recorded on a CMOS camera
with 1024 1024 square pixels, 6.7 m in side. The sample was illuminated by a 532 nm point source located at
17 mm from the center of the CMOS camera. For testing
the self-focusing method, a total of 220 reconstructed images have been utilized in the reconstruction interval
from 3.5 to 5.7 mm, namely, an offset of 10 m, which
is above the limit imposed by the depth resolution of
6.8 m. The results of such comparison for the metrics
utilized for seeking for the best reconstruction plane
of a DLHM hologram of a section of the head of a
drosophila melanogaster fly are shown in Fig. 3. In panel
(a), a plot of the normalized metric versus the
reconstruction distance is presented. Our modified enclosed energy metric reports a global minimum at a
reconstruction distance of 4.2 mm from the point source.
The evaluated metrics show a global critical point around
the reconstruction distance of 4.9 mm. For the evaluated
metrics, we have chosen experimentally a minimum offset of 10 m to match that utilized for our proposed
method. The corresponding reconstructed images for
these distances are shown in panels (b) and (c), in that
order. The reader can see that the image reconstructed in
panel (b) shows details that are barely or not visible at
all in panel (c). The green ellipse bounds an area in which
a spike-like structure is visible in panel (b) but not in
(c). The red ellipse, two-time magnified, surrounds

Fig. 3. Performance of the modified enclosed energy compared with other self-focusing criteria. (a) Normalized metrics
versus reconstruction distance for a DLHM hologram of a section of the head of a fruit fly. (b) and (c) Illustrated reconstructed image for the focal plane found according to the
modified enclosed energy/Tamuras metric.

2572

OPTICS LETTERS / Vol. 39, No. 9 / May 1, 2014

micrometer-sized details that are clearly visible in panel

(b) and totally blurred in panel (c). The performance of
our modified enclosed energy method for self-focusing
DLHM holograms shows that the consideration of the
variation of the irradiance as spherical illumination is
utilized is mandatory. To the best knowledge of the
authors, this is the very first time that a method for
self-focusing with these features is presented; its superior
performance is validated in Fig. 3.
In summary, we have presented an automatic procedure that allows the self-focusing of biological specimens as they are imaged with DLHM. The proposed
method looks for the global minimum of the side of
the square that encloses a given amount energy in an area
surrounding the specimen to be focused. The already
proposed methods in the literature that consider enclosed energy fail when they are applied to DLHM.
Our modified enclosed energy method accounts additionally for the change in the enclosed irradiance, which is
introduced by the propagation of the spherical wavefront
that illuminates the sample in DLHM. The contrasting of
the performance of he proposed method for self-focusing
digital DLHM holograms with some techniques reported
in the literature for different types of specimens leads to
the conclusion that, depending on factors such as the
amount of noise of the images, the object to be focused,
the numerical aperture of the microscope, among others,
diverse metrics can yield different results. The main contribution of our proposal is to provide more robust
results in the self-focusing of holograms recorded via
DLHM than with other self-focusing techniques. Even
though our method has a higher computational cost than
the regular enclosed energy approaches, it keeps the
elapsed time in the frame to be technologically attractive.

The modified enclosed energy method has been validated

by searching for the best focal plane of experimentally
recorded holograms of a paramecium specimen and a
section of the head of a drosophila melanogaster fly.
This work was supported in part by Universidad
Nacional de Colombia grant Hermes code 19384.
References
1. F. C. A. Groen, I. T. Young, and G. Ligthart, Cytometry 6, 81
(1985).
2. T. T. E. Yeo, S. H. Ong, Jayasooriah, and R. Sinniah, Image
Vis. Comput. 11, 629 (1993).
3. F. Dubois, C. Schockaert, N. Callens, and C. Yourassowsky,
Opt. Express 14, 5895 (2006).
4. P. Langehanenberg, B. Kemper, D. Dirksen, and G.
von Bally, Appl. Opt. 47, D176 (2008).
5. P. Langehanenberg, G. Bally, and B. Kemper, 3D Res. 2, 1
(2011).
6. P. Memmolo, C. Distante, M. Paturzo, A. Finizio, P. Ferraro,
and B. Javidi, Opt. Lett. 36, 1945 (2011).
7. P. Ferraro, G. Coppola, S. De Nicola, A. Finizio, and G.
Pierattini, Opt. Lett. 28, 1257 (2003).
8. J. C. Filipinas and P. F. Almoro, Opt. Lett. 36, 3798 (2011).
9. J. F. Restrepo and J. Garcia-Sucerquia, Opt. Lett. 37, 752
(2012).
10. D. Gabor, Nature 161, 777 (1948).
11. J. Garcia-Sucerquia, W. Xu, S. K. Jericho, P. Klages, M. H.
Jericho, and H. J. Kreuzer, Appl. Opt. 45, 836 (2006).
12. J. Garcia-Sucerquia, Appl. Opt. 52, A232 (2013).
13. M. H. Jericho and H. J. Kreuzer, in Coherent Light Microscopy, P. Ferraro, A. Wax, and Z. Zalevsky, eds. (Springer,
2011), pp. 330.
14. M. Born and E. Wolf, Principles of Optics, 7th ed.
(Cambridge University, 2005).
15. M. F. Toy, S. Richard, J. Khn, A. Franco-Obregn, M. Egli,
and C. Depeursinge, Biomed. Opt. Express 3, 313 (2012).