Professional Documents
Culture Documents
We report an experimental implementation of long-range polarimetric imaging through fog over kilo-
metric distance in real field atmospheric conditions. An incoherent polarized light source settled on a
telecommunication tower is imaged at a 1.3 km distance with a snapshot polarimetric camera includ-
ing a birefringent Wollaston prism, allowing simultaneous acquisition of two images along orthogonal
polarization directions. From a large number of acquisitions datasets and under various environmental
conditions (clear sky/fog/haze, day/night), we compare the efficiency of using polarized light for source
contrast increase with different signal representations (intensity, polarimetric difference, polarimetric
contrast,...). With the limited-dynamics detector used, a maximum fourfold increase in contrast was
demonstrated under bright background illumination using polarimetric difference image.
OCIS codes: (110.0113) Imaging through turbid media; (110.5405) Polarimetric imaging;
(010.7295) Visibility and imaging; (110.4280 Noise in imaging systems).
http://dx.doi.org/10.1364/XX.99.099999
state of polarization of light (linear or circular) ing and the corresponding experimental techniques.
during propagation through thick scattering media, Then, in Section 3, we detail the long-range polari-
such as heavy fog conditions, with minor depolar- metric imaging setup used in this experiment, as
ization. This so-called polarization memory effect well as calibration procedures in Section 4. Two
has been analyzed in numerous references, both for representative datasets are then extensively ana-
linearly polarized [17] and circularly polarized light lyzed in Section 5, among numerous datasets ac-
[18, 19]. As a result, such effect can be exploited quired during experimental campaigns. The effi-
using a polarization-sensitive imaging device to en- ciency of different polarimetric signal representa-
hance the visibility of a source provided this latter tions to increase visual contrast of a polarized light
emits polarized light. These approaches have al- source in fog over long distance is discussed for vary-
ready proved efficient and been reported in a num- ing environmental conditions. Then, a general dis-
ber of in laboratory experiments [12–16], with sim- cussion and conclusion on the experiment is even-
ulated turbidity conditions on very short distances tually provided in Section 6.
using artificial scatterers like aerosols.
2. Polarimetric contrast imaging
In case of light travelling through fog, photons Polarization sensitive imaging has proved efficient
undergo scattering by a cloud of randomly dis- in the context of enhanced vision through turbid
tributed particles with sizes in the order of 5-50 media [22], industrial quality control [23, 24] and
µm and varying number density [20]. Moreover, de- machine vision [25]. To probe the complete polari-
pending on the environmental conditions, the scat- metric properties of a light source, one needs to
terers in the intervening medium may not retain a measure the Stokes vector S given by
perfect spheroid shape due to drag (as in case of
large rain drops) or may have ice crystals with pre- S0 Ix + Iy
ferred orientations [21]. S Ix − Iy
S = 1 = (1)
Although these overall studies on polarized light S2 I+45◦ − I−45◦
propagation bring very useful elements of under- S3 IR − IL
standing, they remain quite difficult to exploit for
from which the degree of polarization (DOP) of
sizing an outdoor imaging system because they are
the sourcepcan be obtained using the relation,
carried out in well-established and well-controlled
conditions. Indeed, the combined effects of sun il- DOP = S12 + S22 + S32 /S0 . Hitherto, various
lumination, surrounding artificial illumination, vis- techniques have been employed to fully or partially
ibility evolution, change of scatterers density and measure the Stokes vectors of an image and obtain
size, varying atmospheric conditions are extremely the polarimetric information of the scene of inter-
difficult to mimic or anticipate in a laboratory. est. Most often, rotating polarizers and/or mov-
In addition, unless resorting to numerical scatter- ing birefringent plates are used. Other schemes in-
ing simulations, long range propagation is a phe- cluding prisms [26], Savart plates [27], polarization
nomenon that cannot be easily simulated by a lab- gratings [28], liquid crystal modulators [29] or mi-
oratory experiment, even with a scaling approach. crogrid division-of-focal-plane polarimetric imagers
[30] have been used with varying degrees of com-
Keeping in mind the aforementioned applications promise towards mechanical reliability and real-
and requirement of study of such vision systems in time acquisition and processing. Measuring the full
real atmospheric conditions, we report in this pa- Stokes vector at each pixel can be a slow and storage
per the design, implementation and operation of an heavy task and hence not very suitable for imaging
imaging experiment we developped to investigate moving objects, thus, limiting their application in
the polarimetric contrasts of a scene including a po- real-time scenarios. In the experiment presented
larized light source in real foggy environment over here, we consider a highly polarized source with
a kilometric distance. Such distance corresponds to a priori known linear polarization state with the
a reasonable range requirement for transport safety intervening medium being non-birefringent. There-
applications. We also analyze the efficiency of sev- fore, it is not necessary to measure the full Stokes
eral representations of the polarimetric informa- vector but only the first two components of the
tion obtained with our imaging installation in var- Stokes vector to define the so-called Orthogonal
ious environmental conditions (clear sky/fog/haze, States Contrast given by Eq. (2)
day/night).
This article is organized as follows: in the next S1 Ik − I⊥
OSC = = , (2)
section, we recall some basics of polarimetric imag- S0 Ik + I⊥
3
N
Reception site TDF Tower
1.27 km
Map data © 2014 Google, INEGI
Fig. 1. Long range polarimetric imaging experimental setup. The source and the camera are separated by 1.27 km
with the camera placed in the University of Rennes 1 campus and the source located on a telecommunication tower
of the TDF company. The photograph shows the polarized light source settled on the telecommunication tower.
where Ik and I⊥ are the intensities obtained TDF company. The detection setup is located in
through orthogonally aligned analyzers or through a laboratory building as illustrated in Fig. 1. The
a polarization splitting Wollaston prism. In the various parts of the setup are described in detail in
case considered here of a linearly polarized light the following subsections.
propagating through non-birefringent medium, the
OSC is equal to the DOP of the source after pass-
ing through the atmosphere and fog. The inten- 3.A. Polarized Source
sity measures recorded by a pixel of a camera are The source used is a 300 W halogen incandescent
of course affected by various noise sources, for in- lamp with a linear polarizer, both placed inside a
stance, (a) the Gaussian electronic noise induced weatherproof steel housing as shown in Fig. 1. The
by the electronic read-out circuit of the CCD array light from the incandescent lamp is polarized us-
sensor ; and (b) optical noise due to photon noise, ing an adhesive dichroic polymer polarizing sheet
atmospheric turbulence and fluctuations introduced glued to a glass plate and oriented such that the
by spatial/temporal evolutions of the intermediate polarization axis is vertical. Half the optical power
foggy medium. As will be seen later, the com- provided by the lamp being absorbed by the polar-
putation of the OSC may be strongly affected by izer, it turned out that standard polaroid sheets are
this noise, especially in low illumination conditions, subject to deterioration after a few minutes of illu-
since the division by total intensity (IT = Ik + I⊥ ) mination. For this reason, we use polarizing sheets
may lead to unbounded values. specifically designed for LCD projector industry,
thus ensuring high durability under high power op-
3. Long range Experimental Setup eration conditions. Before installing the source, we
The experiment presented here is designed to imple- conducted durability tests in indoor conditions with
ment a long-range polarimetric imaging system over a 300 W lamp. The lamp was allowed to run con-
kilometric distance, and in real-field outdoor condi- tinuously during day times. In these conditions, it
tions. Such experiment allows us to assess the bene- was noticed that the polarimetric contrast of the
fits of using polarized light for improved detectabil- sheet was reduced by 57.4% in a period of 5 days
ity of a light-mark through turbid atmosphere over with continuous daytime usage. In light of this, a
long distances. Such a system can have practical weather-proof steel housing was designed, such that
applications in aviation or navigation at sea in poor the polarizing sheet glued on a glass plate is placed
visibility conditions. The setup mainly consists of sufficiently far (20 cm) from the lamp. Moreover,
a polarized light emitter and a polarimetric imag- holes have been drilled in the bottom and top of
ing system that simultaneously acquires two images the housing to enable air-flow in-between the lamp
of a scene corresponding to the two orthogonal di- and the glass plate. This air flow is stimulated by
rections of polarization. These elements, which are convection when the temperature of the glass-plate
extensively described in this section, are located on increases allowing the hot air in-between to be re-
the Beaulieu campus of University of Rennes 1, and moved, and thus providing an efficient heat dissipa-
are separated by 1.27 km. The source is located on tion. Such mechanical design ensures a reasonable
the top of a telecommunication tower owned by the lifetime of almost one year in outdoor conditions
4
before requiring replacement. ation of 0.6 e− /s. The camera was selected for its
The telecommunication tower is about 80 m in low noise properties in low light conditions, which
height and provides a suitable location for the as we will see, is preferable while performing arith-
source. The distance of the tower from the labo- metic on the acquired polarimetric images.
ratory site in the University of Rennes 1 is optimal
for various applications where a long-range imaging
through fog may be crucial. The tower is also the
tallest structure in the surrounding area and is in
line-of-sight from laboratory buildings. The polar-
ized source is connected to a GSM switch enabling
easy remote control of the emission part of the ex-
periment.
overall scene. The advantage of this mode is most 4.A. Imaging setup calibration
apparent during twilight and in foggy conditions The frames obtained in the experiment described
when the illumination of the scene varies strongly above are binary values obtained from the cam-
in time. Using dynamically controlled exposure era. Processing of the frames to extract the im-
time, one obtains a time series of frames, which can ages is at this step carried out in a post-acquisition
be normalized to their respective exposure times stage. The post-processing includes fixed pattern
(which are stored in a data-file by the program) so dark frame subtraction, pixel non-uniformity cor-
as to continually exploit the full dynamic range of rection, image registration and optical distortion
the camera. We also implement a so-called cumula- correction. The latter processing steps are orga-
tive grab mode (CGM) to avoid recording the 100 nized in the following manner.
Hz intensity fluctuations due to the 50 Hz modula-
tion of the electrical supply network. This mode is Detector calibration: For a camera sensor
automatically activated for exposures smaller than there are mainly three calibrations to be performed
10 ms, that is when the source blinking becomes in order to record an image that closely corresponds
apparent. Under CGM mode, a frame having max- to the scene being imaged, namely dark pattern cal-
imum gray-level value of the source pixel is chosen ibration, non-uniformity of pixel gain and removal
from a sample of NCGM frames where NCGM de- of hot and dead pixels. Each pixel has different
pends on the exposure time Texp in milliseconds as dark noise properties at a particular temperature
NCGM = 20(ms)/Texp . The program also displays and exposure time. After averaging over a number
the calculated OSC of the source pixel in real-time of dark frames acquired in total darkness, one ob-
and allows for compensation of the ambient light tains the fixed pattern dark frame (IF P D ) that can
contribution, in which case the OSC reads be subtracted from the images of interest. It must
be noted that this correction depends on the expo-
s − Ia )
(Iks − Ika ) − (I⊥ sure time. We thus averaged 500 dark frames to
⊥
OSCA = . (3) obtain the IF P D for different exposure times. An-
(Iks − Ika ) + (I⊥
s − Ia )
⊥ other important correction to be made is the non-
uniformity in pixel gain. Owing to the fabrication
In the above equation, Iks (respectively I⊥ s ) denotes process, each pixel in the sensor may not have the
the average intensity over a 3 × 3 pixels area enclos- same sensitivity/gain. This non-uniformity is es-
ing the pixel corresponding to the source location timated by illuminating the sensor with spatially
(source pixel) in Ik (respectively I⊥ ). On the other uniform white light and recording the resulting pat-
hand, Ika (respectively I⊥a ) estimates the ambient tern. This image is normalized by dividing by the
illumination by averaging a 3 × 3 pixels area in mean gray level to create a gain non-uniformity im-
the vicinity of the source pixel, but strictly distinct age IGN U . Using the above two calibration images,
from the source spatial extent. With such defini- a raw image recorded is systematically corrected
tion, the OSC is computed taking into account in- using Icorr = (Iraw - IF P D )/IGN U .
tensity measures Ika and I⊥a of a reference area close Image registration and distortion calibra-
to the source representing the ambient illumination. tion: Once performed the above calibrations and
Using such control software with AEM allows us corrections on the bare detector itself (without
to acquire images at predefined intervals over long any image forming optics), the polarimetric optical
periods of time with strongly varying illumination imaging system is mounted. When the system is
conditions (i.e., clear sky, obstructed vision, day- assembled and images are recorded, one has to ex-
time/night-time and twilight). tract the two images corresponding to orthogonal
polarization directions from a single frame. It is re-
quired that the two extracted images have a one to
4. Experiment calibration one correspondence between pixels such that they
In this section, we first describe the camera cal- map the same scene. This process is straightfor-
ibration and image registration procedures which ward if there is no geometrical distortion or if both
have to be implemented so as to provide reliable the images share the same geometrical distortion.
polarimetric images and measurements. Then, we However, the WP used in the polarimetric imaging
report some experimental results which allowed us setup does introduce astigmatism and anamorphic
to validate the experimental system (emission and distortion in the resulting images [31]. The distor-
acquisition) and laboratory calibration on real field tion is non-symmetrical with respect to the two im-
conditions. age channels and hence a perfect image registration
6
around the source location, we will limit ourselves imaging system can be aligned in line-of-sight with
to simple translation registrations, for the sake of the source on the tower. In addition, the tilt angle
computational effectiveness on long time series ex- (roll) must be well aligned so that the polarization
periments. axis of the imager is strictly aligned with the po-
Intensity Asymmetry: It is important to larization direction of the source. This calibration
take note of any transmission asymmetry in the is preferably done during night time in clear sky
two image channels introduced by the WP. In our with the polarized lightmark turned on. It is in-
case a 2.7 % asymmetry in intensity was observed deed easier to minimize and even null the intensity
after imaging a uniformly illuminated depolarizing at the source location in the perpendicular polariza-
object (plain white paper observed with different tion image (I⊥ ) in dark background. After such cal-
orientations) through the system. The measured ibration, we observed an OSC of 0.97 for the source
average intensity ratio (Ik /I⊥ = 1.027) was used to pixel which increased to a value of 1.00 when OSC
correct this asymmetry. This ratio was seen to be was computed taking into account the low intensity
constant over a large central region including the background contribution, as in Eq.(3).
region of interest. During these preliminary calibration acquisitions,
it was checked on the acquired frames that the day-
time clear sky is polarized to some extent, depend-
ing on sun location, and that the ambient light
100
backscattered on the tower structures and imaged
200 on the camera can partly retain polarization. How-
300 ever, we observed that the imaged scene is almost
400
totally depolarized as soon as a thin layer of fog
or haze is present in the atmosphere, or when the
500
weather is covered by (even very thin) clouds. As a
600 result, the difference image of the two image chan-
700 nels is expected to provide an obvious advantage
over intensity imaging due to its ability to cancel
100 200 300 400 500 the unpolarized background and isolate the polar-
ized source.
Fig. 4. The raw image obtained from the camera with 5. Experimental results: Signal contrast in
782 × 582 pixels. Extraction of two image channels after foggy condition
applying image registration process is demarcated in yel- Once settled, optimized and calibrated the imag-
low dashed box. The ambient illumination is obtained
ing system described in previous sections, and us-
by averaging over the region shown in red-dashed lines
ing the above mentioned control program, we were
and the region shown in blue dashed lines is the 21 × 21
able to acquire numerous series of frames of the
pixels ROI that includes the source at its center.
source at different times of the day (dawn, day-
time, dusk, nigthtime) with varying atmospheric
4.B. In-situ verification conditions (clear sky, thin haze, fog, cloud, rain,
Along with the above calibration procedures, fur- snow,...). The corresponding time-series acquisi-
ther alignment of the detection system with the tions were recorded during years 2011 and 2012
polarization direction of the source is done while with various time periods (10 images/minute to 6
mounting the detection system. The polarimetric images/hour) over long time spans of several hours
camera setup and the control/acquisition computer or tens of hours.
are settled in a mezzanine floor of a building fac- During these long term acquisitions, the dynam-
ing the telecommunication tower. It has first been ically varying exposure time in AEM was recorded
checked that the spread of the light source was close for each acquisition and the frames were normal-
to our expectancies (1 × 1 pixel with the selected ized to their respective exposure times during post-
50 mm focal length objective lens), with an actual acquisition analysis. From each raw frame, we then
horizontal spread of 1.5 pixels and vertical spread extracted the two sub-images Ijk and Ij⊥ (with j cor-
of about 2 pixels. The main discrepancy in the ver- responding to the j th image in the time series). One
tical direction is due to the influence of the WP. raw image extracted from a time series is repre-
The whole setup is placed on a stable mechani- sented in Fig. 4, with an overlay indicating the im-
cal mount with adjustable angle (pitch) so that the age registration (rectangles in yellow) from which
8
Signal representation Symbol Definition we will be interested in the remainder of this sec-
Intensity γI Ik + I⊥ tion in comparing the visual contrast of the polar-
Polarimetric difference γ∆ Ik − I⊥ ized source that can be obtained with these differ-
Ik −I⊥
OSC γOSC Ik +I⊥ ent representations. We must therefore provide a
Polarimetric ratio γratio
Ik fair comparison between the contrasts obtained in
hI⊥ iχk
different representations. For this purpose we de-
Table 1. Signal representations of polarimetric data, fine a contrast-to-noise ratio (CNR) such that the
with hiχk denoting spatial averaging over a square neigh- CNR value is insensitive to the scale of these four
bourhood χk of the central pixel of size Nk pixels. representations. The CNR is defined for a central
pixel denoted s (signal γs ) with respect to its sur-
rounding square neighbourhood χk of size Nk pixels
regions of interest (ROI) corresponding to Ijk and as,
Ij⊥ are extracted. The example of ROI shown in γs − hγiχk
light blue in Fig. 4 is a 21 × 21 pixels rectangle CN R = q 2 (4)
1 P Nk
surrounding the source location. Nk −1 i∈χk γi − hγiχk
P Nk
5.A. Principle of the measurement where hγiχk = i=1 γi /Nk is the spatial average
These numerous acquisitions were used to gauge the over χk . It must be noted that the evaluation of
effectiveness of polarimetric imaging under various the spatial averages are carried out for a sample of
atmospheric conditions. The analysis of the exper- pixels defined by the region χk that excludes the
imental results led us to retain four main different 2 × 2 central region centered on the source pixel to
scenarios, associating clear sky / foggy conditions prevent spatial extent of the light source from per-
and day/night time. The conditions in daytime turbing computation of the background statistical
with fog are indeed such that the intensity of the properties. With the above definition, the CNR is
pixels around the source are nearly uniform, mak- basically the measure of how many standard devia-
ing it difficult to identify the source from the back- tions away does the signal pixel lie from the distri-
ground in a simple intensity image. In clear sky bution of its surrounding pixels. The CNR values
conditions, the background consists of the telecom- computed on the data provide a more quantitative
munication tower structures and installations. On evaluation of the contrast than a visual analysis of
the other hand, in night time we have conditions the polarimetric images plotted in subsequent fig-
where the source is brighter and the background ures. Moreover, in an applicative context, the CNR
is dark with unpolarized and localized bright light would be directly linked to the performance of au-
sources (for instance, the red aircraft warning lights tomated detection of polarized sources of unknown
on the tower) or sometimes, during heavy fog, a uni- number and locations in a scene. In the presence of
formly lit background because of scattering of city structured background and/or unpolarized spurious
lights. sources, a high CNR would warrant good detection
In all these conditions, we are interested in maxi- probability along with low false alarm rate.
mizing the visual contrast of the source with respect In the following subsections, we provide a thor-
to its background. Polarimetric imaging provides ough data analysis performed on two particular
us with two data sets (Ik and I⊥ ) for the same image time-series acquisitions, allowing one to observe
and thus allows us to create various representations clear sky and foggy conditions in day time (Subsec-
of the image with different visual contrasts. We tion 5.B) and night time (Subsection 5.C). These
thus propose to consider four simple signal repre- two experimental data series are representative of
sentations of the available data, generally referred the wide variability of environmental conditions
to as γ, and compare their visual contrast under that were made possible to address with the ex-
various environmental conditions. These four signal perimental facility described in this article.
representations correspond to a simple intensity im-
age (γI ), a polarimetric difference image (γ∆ ), the 5.B. Day time measurement
OSC image (γOSC ) and a polarimetric ratio image This first data acquisition series that we will focus
(γratio ), and are given in Table 1, along with their on has been recorded on the 17-Oct.-2011, between
respective symbol and mathematical definition. 10.02 am and 10.57 am. During this 55-minutes
It is obviously seen that the absolute values of acquisition, an image has been recorded every 10
these signal representations are different. However, seconds, making it possible to observe rapid rising
9
100
Jratio
[b]
20
0
[H] [I] [J] [K]
10:05:13 10:15:14 10:25:14 10:35:18 10:45:18 10:55:18
Time (hh:mm:ss)
Fig. 5. (a) Time evolution of the source (signal) intensity, ambient intensity and camera exposure time across day
time experiment. (b) Time-evolution of the CNR of four polarimetric signal representations. (c) Raw images of four
frames labelled H, I, J and K, corresponding to various visibility conditions across day time experiment.
indicated in Fig. 6 correspond to the CNR of the ments of the telecommunication tower can be iden-
central pixel estimated on this neighbourhood. As tified on the intensity image. Indeed, as it has been
can be seen from these results, the source is hid- noticed in Subsection 4.B, the ambient light does
den behind heavy fog until 10.13 am and all CNR not exhibit noticeable polarization as soon as the
values remain close to zero in this situation. It can sky is covered by clouds or when a thin layer of haze
be indeed checked in Fig. 6 that the source is not or fog is obscuring the atmosphere. Hence, the con-
distinguishable on any of the four signal represen- tributions of this ambient background on the two
tations studied. polarimetric images I// and I⊥ are strictly identi-
Measurements performed after 10.13 am are more cal, and can be suppressed on a difference image.
interesting. Indeed, they first allowed us to observe Lastly, it is interesting to note that γOSC never
that, as soon as the light emitted by the source is appears to be the best representation, contrary
visible over the background, it retains a very high to anterior work on active polarimetric imaging
polarization degree, with no detectable depolariza- [33, 34] (where, due to the reflection geometry of
tion induced by turbid atmosphere, nor partially imaging, the background has a significant polar-
depolarized halo in the vicinity of the source loca- ized component). Indeed, in the present context
tion. Following a similar interpretation as in prior of active polarized source and passive totally unpo-
laboratory experiments [12–17] on short distances, larized background, OSC implies a normalization
this result can be understood considering proper- of γ∆ by the total intensity γI , which, on the dark
ties of Mie scattering on large particle size in fog background image γ∆ obtained, increases the rela-
droplets [16, 32]. In our case, since we are work- tive noise in the final image and thus decreases the
ing at a very long distance, the light emitted by the CNR. In the next subsection, it will be shown that
source can be almost totally absorbed by a thin fog, such normalization can be much more detrimental
which in return cannot be considered as a strongly to signal quality in night time conditions. There-
diffusive medium, and hence cannot induce strong fore, this normalization must be avoided when it
depolarization. In addition, the low numerical aper- comes to locating or detecting a target in poor vis-
ture of the imaging system at such distance acts as a ibility conditions, unless one wants to assess the
spatial filter [6, 7], thus preventing most of the scat- DOP of the light source.
tered and snake photons (slightly scattered) from
entering the imaging system. 5.C. Night time measurement
Then, the results given in Figs. 5 and 6 reveal A similar experiment was conducted during night
that, as soon as intensity coming from the source is time on the 14-Oct.-2011, between 1.53 am and
detectable, all polarimetric signal representations 7.17 am. During this 324 minutes acquisition, a
(γ∆ , γOSC and γratio ) provide CNR values sig- raw frame has been recorded every minute. As in
nificantly higher than the CNR obtained with a the previous subsection, the evolution of signal pixel
standard non-polarimetric intensity measurement and background intensities are plotted respectively
(γI ). This general result thus confirms the inter- in red and black curve in Fig. 7.(a). The evolution
est of polarimetric measurements for enhanced vi- of the exposure time over this long time period is
sion through poor visibility conditions. The analy- given in Fig. 7.(a) in dotted blue curve (logarith-
sis can be however refined so as to identify the best mic scale). This curve shows the strong variations
signal representation in a given situation. In very of the exposure time at night with atmospheric con-
low visibility conditions (acquisition I), it appears ditions, ranging from 7 ms to a maximum value of
that γratio gives the best improvement in terms of 4 s authorized by the camera used. From the obser-
CNR (60% increase) with respect to an intensity vation of these curves, it can be noted that a dense
measurement γI . fog cloud has totally obscured the visibility of the
On the other hand, when visibility increases, the source between 3.40 am and 4.55 am, leading to
best signal representation corresponds to a differ- a maximum exposure time in total darkness, as in
ence image γ∆ allowing the CNR to be enhanced situation labelled M in Fig. 7. On the other hand,
up to a factor of almost 4 (acquisition K). The ef- the beginning and end of the acquisition correspond
ficiency of the difference image is due to its ability to almost clear sky conditions (as in situations la-
to suppress the background contribution. This en- belled L and O). The strong signal intensity vari-
hancement is either due to diffused light in the tur- ations observed in clear sky conditions are mainly
bid atmosphere, as in acquisition J, or due to the due to atmospheric turbulence which causes flicker
surrounding scene, as in acquisition K on which ele- of the location of the signal pixel on the images, and
11
[b]
200
40 [N]
0 20
Fig. 7. Same as Fig. 5 for night time experiment. The inset shows a magnified part of the plot where the fog is
thinning and the CNRs are comparable for all signal representations. Four frames labelled L, M, N and O have been
selected to be representative of visibility conditions during night experiments. The display dynamics in (c) has been
expanded by a factor of 20 for all night situations so that the background becomes visible.
improving the CNR values by a factor comprised sual contrast or detectability of such light source
between 10% and 35%, which can be explained by in degraded visibility conditions, and at a distance
the noise reduction induced by the spatial averaging compatible with potential applications, such as nav-
of the background in the expression of γratio . igation assistance for airborne, maritime or terres-
These results, which seem to soften the inter- trial transportation. It can be concluded from these
est of polarimetric imaging for vision enhancement results that, for the purpose of detecting a polarized
though turbid atmosphere in night conditions, ac- source embedded in an unpolarized background, us-
tually reveal the limitations of the sensor used for ing polarized beacons and polarimetric cameras can
these experiments. These nocturnal acquisitions in- lead to a significant improvement of the contrast.
deed allow us to identify two ways of improvement This is particularly true in the presence of high
for our detection system, which are (a) the mini- background illumination levels, which is likely to
mization of the dark noise, and most importantly occur in daytime conditions, or in night-time sit-
(b) the increase in detector dynamics. Reducing uations, due to scattering of unpolarized ambient
the dark noise would limit the level of noise in light by the atmosphere itself. This is indeed bene-
dark background in night-time conditions for po- ficial, as computing a simple polarization-difference
larimetric signal representations, as stated above. image would make it possible to isolate the target
Moreover, it can be observed in Fig. 7.(b) that from ambient light and hence provide higher con-
between 6.00 am and 7.10 am, strong variations trast in foggy situations in real-time.
in all the CNR values occur, following the same Another result of this article is that no signifi-
abrupt “steps” at each modification of the expo- cant depolarization of the polarized source, nor de-
sure time. This is a confirmation that in these sit- polarized halo could be detected during propaga-
uations, the dark image background is perturbed tion through fog over a kilometer with a standard
with camera dark noise. Usually, the dark noise dynamics camera, as the one used in this exper-
level increases with the acquisition time. How- iment. This observation, unreported to our best
ever, the Basler A312f camera used implements knowledge at a kilometric distance, thus confirms
a Sony ICX 415AL sensor, whose HAD (Hole- the interest of conducting such experiments in real-
Accumulation Diode) technology makes it possible field conditions, over long distances. This behavior,
to inhibit the increase of dark noise level with inte- similar to the one observed on short distance labo-
gration time. As a result, the noise contribution is ratory experiments was not an entirely predictable
thus artificially doubled in our measurements, rela- result. It is indeed difficult to mimic the physi-
tively to the signal pixel value when exposure time cal conditions that were encountered during these
is divided by two, hence leading to a strong decrease experiments with in laboratory short distance mea-
in terms of CNR. This also confirms that adapting surements using highly diffusive water or gas tanks.
the exposure time cannot increase the dynamics of a
detector when the illumination variations across the As a perspective to this work, we are currently
scene exceed the actual sensor dynamics. It would implementing to our setup a new camera with a
thus be highly beneficial to use a high dynamics sen- higher dynamics range and lower noise level, so as
sor to improve these results and potentially increase to refine the first analyses presented in this arti-
the range of environmental conditions for which po- cle. Such a camera will make it possible to ana-
larimetric imaging could be advantageous for vision lyze situations in which observations were limited
enhancement through turbid atmosphere over long by the detector itself. More importantly, it is also
distances. expected that a high dynamics sensor will increase
the range of application of polarimetric imaging in
6. Conclusion terms of visibility distance for beacon detection or
In this paper, we have reported an experimental localization in turbid atmosphere. Other impor-
setup allowing to perform snapshot polarimetric tant perspectives to this work can be enumerated,
contrast imaging of a light emitter through tur- among which the definition of clear figures of merit
bid atmosphere over kilometric distance. With this that link image contrast ratios with visibility range.
setup it has been possible to acquire long time- Another important field of research that will be
series of measurements in various real environmen- addressed is the definition of optimal signal rep-
tal conditions of weather and visibility. The re- resentation as a function of environmental or in-
sults reported in the article have permitted to as- strumental conditions for various image processing
sess the benefits of using a polarized light source tasks, such as estimation or detection. Recent re-
and a polarization-sensitive camera to increase vi- sults have evidenced the potential of polarimetric
13
imaging contrast when relevant statistical parame- [9] L. Mullen, A. Laux, B. Concannon, E. P. Zege, I. L.
ters, such as partial correlation between measure- Katsev, and A. S. Prikhach, “Amplitude-modulated
ments, are considered [35]. Applying these theo- laser imager,” Applied optics 43, 3874–3892 (2004).
retical results to experimental data will necessitate [10] E. Lacot, R. Day, and F. Stoeckel, “Laser optical
a more thorough analysis of the images in terms feedback tomography,” Optics letters 24, 744–746
of statistical properties, which will also be enabled (1999).
with the use of a better camera. Then, the imple- [11] E. Belin and V. Boucher, “An imaging system based
mentation of these optimal estimators/detectors on on laser optical feedback for fog vision applica-
FPGA boards for in-line real-time image processing tions,” in “SPIE Conference Series,” , vol. 7088
and display is also an interesting perspective to this (2008), vol. 7088, p. 70880N.
[12] J. Schmitt, A. Gandjbakhche, and R. Bonner, “Use
work.
of polarized light to discriminate short-path photons
in a multiply scattering medium,” Applied optics
Acknowledgments 31, 6535–6546 (1992).
The authors are grateful to the TDF company, and [13] H. Horinaka, K. Hashimoto, K. Wada, Y. Cho, and
in particular to A. Lecluse and P. Gelebart, for pro- M. Osawa, “Extraction of quasi-straightforward-
viding location for the emitter on the telecommu- propagating photons from diffused light transmit-
nication tower. This work has been partly funded ting through a scattering medium by polarization
by the CEFIPRA (project N◦ 4604-4). modulation,” Optics letters 20, 1501–1503 (1995).
[14] O. Emile, F. Bretenaker, and A. Floch, “Rotating
polarization imaging in turbid media.” Optics let-
References ters 21, 1706–1708 (1996).
[1] S. Andersson-Engels, R. Berg, S. Svanberg, and [15] H. Ramachandran and A. Narayanan, “Two-
O. Jarlman, “Time-resolved transillumination for dimensional imaging through turbid media using a
medical diagnostics,” Optics letters 15, 1179–1181 continuous wave light source,” Optics Communica-
(1990). tions 154 (1998).
[2] Z. Nan, J. Xiaoyu, G. Qiang, H. Yonghong, and [16] S. Mujumdar and H. Ramachandran, “Imaging
M. Hui, “Linear polarization difference imaging and through turbid media using polarization modula-
its potential applications.” Applied optics 48, 6734– tion: dependence on scattering anisotropy,” Optics
6739 (2009). communications 241, 1–9 (2004).
[3] M. Dubreuil, P. Delrot, I. Leonard, A. Alfalou, [17] J. S. Ryan and A. I. Carswell, “Laser beam broad-
C. Brosseau, and A. Dogariu, “Exploring underwa- ening and depolarization in dense fogs,” J. Opt. Soc.
ter target detection by imaging polarimetry and cor- Am. 68, 900–908 (1978).
relation techniques,” Applied optics 52, 997–1005 [18] M. Xu and R. Alfano, “Circular polarization mem-
(2013). ory of light,” Physical Review E 72 (2005).
[4] W. R. Watkins, D. H. Tofsted, V. G. CuQlock- [19] R. Nothdurft and G. Yao, “Applying the polariza-
Knopp, J. B. Jordan, and J. O. Merritt, “Navigation tion memory effect in polarization-gated subsurface
through fog using stereoscopic active imaging,” in imaging.” Optics express 14, 4656–4661 (2006).
“Enhanced and Synthetic Vision 2000, SPIE Con- [20] J. A. Garland, “Some fog droplet size distributions
ference Series,” , vol. 4023 (2000), vol. 4023, pp. obtained by an impaction method,” Quarterly Jour-
20–28. nal of the Royal Meteorological Society 97, 483–494
[5] L. Wang, P. Ho, C. Liu, G. Zhang, and R. Alfano, (1971).
“Ballistic 2-d imaging through scattering walls us- [21] M. Hayman, S. Spuler, B. Morley, and J. VanAn-
ing an ultrafast optical kerr gate.” Science (New del, “Polarization lidar operation for measuring
York, N.Y.) 253, 769–771 (1991). backscatter phase matrices of oriented scatterers,”
[6] Q. Z. Wang, X. Liang, L. Wang, P. P. Ho, and Optics express 20, 29553–29567 (2012).
R. R. Alfano, “Fourier spatial filter acts as a tem- [22] S. Demos and R. Alfano, “Optical polarization
poral gate for light propagating through a turbid imaging,” Applied Optics 36, 150–155 (1997).
medium,” Opt. Lett. 20, 1498–1500 (1995). [23] P. Terrier, V. Devlaminck, and J. Charbois, “Seg-
[7] E. Leith, B. Hoover, S. Grannell, K. Mills, H. Chen, mentation of rough surfaces using a polarization
and D. Dilworth, “Realization of time gating by use imaging system.” Journal of the Optical Society of
of spatial filtering.” Applied optics 38, 1370–1376 America. A, Optics, image science, and vision 25,
(1999). 423–430 (2008).
[8] M. Oleary, D. Boas, B. Chance, and A. Yodh, “Re- [24] O. Morel, C. Stolz, F. Meriaudeau, and P. Gorria,
fraction of diffuse photon density waves,” Physical “Active lighting applied to three-dimensional recon-
Review Letters 69, 2658 (1992). struction of specular metallic surfaces by polariza-
14
tion imaging.” Applied optics 45, 4062–4068 (2006). cal filters,” Optics express 18, 19087–19094 (2010).
[25] M. Yamada, K. Ueda, I. Horiba, and N. Sugie, [31] M. C. Simon, “Wollaston prism with large split an-
“Discrimination of the road condition toward un- gle,” Appl. Opt. 25, 369–376 (1986).
derstanding of vehicle driving environments,” Intel- [32] D. Bicout, C. Brosseau, A. Martinez, and
ligent Transportation Systems, IEEE Transactions J. Schmitt, “Depolarization of multiply scattered
on 2, 26–31 (2001). waves by spherical diffusers: influence of the size
[26] A. Bénière, M. Alouini, F. Goudail, and D. Dolfi, parameter,” Physical Review E 49, 1767 (1994).
“Design and experimental validation of a snapshot [33] M. Alouini, F. Goudail, N. Roux, L. Le Hors,
polarization contrast imager.” Applied optics 48, P. Hartemann, S. Breugnot, and D. Dolfi, “Ac-
5764–5773 (2009). tive spectro-polarimetric imaging: signature mod-
[27] H. Luo, K. Oka, D. Edward, M. Kudenov, eling, imaging demonstrator and target detection,”
J. Schiewgerling, and E. Dereniak, “Compact and The European Physical Journal Applied Physics 42,
miniature snapshot imaging polarimeter.” Applied 129–139 (2008).
optics 47, 4413–4417 (2008). [34] A. Bénière, F. Goudail, D. Dolfi, and M. Alouini,
[28] M. W. Kudenov, M. J. Escuti, N. Hagen, E. L. Dere- “Target detection in active polarization images per-
niak, and K. Oka, “Snapshot imaging mueller ma- turbed with additive noise and illumination nonuni-
trix polarimeter using polarization gratings,” Opt. formity,” JOSA A 26, 1678–1686 (2009).
Lett. 37, 1367–1369 (2012). [35] J. Fade, S. Panigrahi, and M. Alouini, “Optimal
[29] S. Alali, T. Yang, and I. Vitkin, “Rapid time-gated estimation in polarimetric imaging in the presence
polarimetric stokes imaging using photoelastic mod- of correlated noise fluctuations,” Optics Express 22,
ulators.” Optics letters 38, 2997–3000 (2013). 4920–4931 (2014).
[30] V. Gruev, R. Perkins, and T. York, “Ccd polariza-
tion imaging sensor with aluminum nanowire opti-