You are on page 1of 9

Journal of Cultural Heritage 8 (2007) 151e159

http://france.elsevier.com/direct/CULHER/

Original article

2-D image analysis: A complementary tool for characterizing quarry


and weathered building limestone
Olivier Rozenbaum*, Emmanuel Le Trong1, Jean-Louis Rouet2, Ary Bruand3
Institut des Sciences de la Terre d’Orléans (ISTO), UMR6113 CNRS/Université d’Orléans, 1A rue de la Férollerie, 45071 Orléans Cedex 2, France
Received 6 July 2006; accepted 19 January 2007

Abstract

The understanding of weathered processes and, more generally, of transfer properties of building stones, requires a detailed knowledge of
porosity characteristics. This study aims to analyze two-dimensional (2-D) images of stones by using mathematical tools that enable the descrip-
tion of the pore and solid phase distribution. We selected two limestones that have been widely used for different types of buildings: a quarried
and weathered Tuffeau stone (the latter being used in most châteaux of the Loire) and a quarried Sebastopol stone selected for numerous build-
ings in Paris. Backscattered electron scanning images obtained on thin sections of the stones were studied by using autocorrelation function
analysis and chord distributions. Results showed that these mathematical tools are able to discuss, quantitatively and statistically, differences
of pore and solid distributions between quarried limestones, and to discuss the degree of weathering of stones collected from buildings.
Thus, very small differences of pore and solid phase distribution between the samples studied were revealed by chord distribution analysis
and autocorrelation function analysis. Resulting characteristics obtained with such an analysis are promising information for a better understand-
ing of weathering mechanisms.
Ó 2007 Elsevier Masson SAS. All rights reserved.

Keywords: Limestone porosity; Image analysis; Segmentation; Chord length distribution; Autocorrelation function

1. Research aims phase characteristics of the stone in depth (from a few micro-
metres down to a few centimetres), leading to its irreparable
External environmental agents, such as rain, wind, temper- destruction. Depth and weathering facies depend both on
ature and pollution are the main factors in stone weathering environment factors and stone characteristics [2]. Besides,
[1e3]. Pollutants such as particulate matter originating from a more complete understanding of weathering mechanisms
industry and vehicle exhaust (combustion of oil-derived fuels) requires one to relate the microstructure characteristics to the
and SO2 combined with wetting/drying cycles are responsible macroscopic properties (e.g. permeability, friability). Thus, it
for the alteration by sulphating building limestone [3e7]. In- is important to characterize stones originating from carriers
deed, the acid attack of these geomaterials leads to dissolution and weathered buildings as an initial step for other studies
of carbonates and formation of sulphate compounds (mainly in the field of building stone conservation.
gypsum). The result is a radical change in the porous and solid As water transfer and mechanical properties are linked to
pore network characteristics, the main objective of this paper
is the morphological and structural characterization of the
* Corresponding author. Tel.: þ33 2 38 25 52 44; fax: þ33 2 38 63 64 88. stones. Some pore networks, such as those of stones, can be
E-mail addresses: rozenbaum@cnrs-orleans.fr (O. Rozenbaum), manu@ represented in a first approach by small volumes (pores) con-
mixtion.org (E. Le Trong), jean-louis.rouet@labomath.univ-orleans.fr (J.-L. nected by narrow pores. The minimum section which connects
Rouet), ary.bruand@univ-orleans.fr (A. Bruand).
1
Tel.: þ33 2 38 25 53 99; fax: þ33 2 38 63 64 88. two pores is called a throat. Some classical experiments enable
2
Tel.: þ33 2 38 25 57 27; fax: þ33 2 38 63 64 88. pore phase characterization of disordered porous media, such
3
Tel.: þ33 2 38 25 53 98; fax: þ33 2 38 63 64 88. as those of stones: mercury intrusion porosimetry (e.g.

1296-2074/$ - see front matter Ó 2007 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.culher.2007.01.004
152 O. Rozenbaum et al. / Journal of Cultural Heritage 8 (2007) 151e159

[8,9]), nitrogen BET (e.g. [10]) or image analysis formed during the middle Lutetian (45  106 years). It was
(e.g. [11,12]). It should be kept in mind that the first two tech- extracted to build historic buildings in Paris. The Sebastopol
niques measure a pore throat size distribution depending on stone studied in this paper was mainly composed of 81.6% cal-
the cross-sectional throat shape and topology of the pore throat cite and 16.8% silica (determined by induced chemical
network. These techniques give on first approximation an plasma). Total porosity was 43.4% but with a pore size
equivalent circular cross sectional diameter by way of a geo- distribution corresponding to larger pores than those forming
metrical model, giving inaccurate information on the pore size, porosity of the Tuffeau stone. Indeed, the equivalent pore
even if they produce a satisfactory first approach of the pore size distribution ranged from 0.5 to 200 mm in size [18].
network. In this study, we present other tools based on image A weathered Tuffeau stone was also selected. It originated
analysis that enable to us to improve the knowledge of the from the Saint-Donatien church located in Orléans (France).
solid and porous phase distribution of a stone. Within the last Blocks were extracted from the north-east wall of the bell tower
20 years, analyses of images produced by using optical micro- at an elevation of 20 m. Sampling occurred during the church
scopy or scanning electron microscopy (SEM) were commonly restoration that consisted of replacement of whole blocks by
used to characterize different porous materials, such as porous news ones. The exposed surface had a grey crust, which was
silica, soils, concrete, stones [13,14]. The present work focuses harder than the stone core and the first 3 cm underneath that sur-
on computerized 2-D image analysis using mathematical tools face were crumbly with micro-cracks parallel to the surface. A
characterizing geomaterials in different ways. block was selected and, as the weathering degree decreased
In this paper, we selected two quarry limestones used as with depth from the surface, the stone was sampled in the 0e
building stones and a weathered limestone originating from 55 mm range from the surface. Chemical analyses (induced
a church. The aim is a quantitative characterization of the chemical plasma and infrared spectroscopy) confirmed that
pore and solid phases of the selected stones by using mathemat- this stone was mainly composed of calcite and silica, like the
ical tools applied to image analysis. The present samples were quarry Tuffeau stone [16,17]. Infrared spectroscopy and micro-
selected because they have been widely studied earlier by clas- probe analysis showed that gypsum was present in the first
sical techniques [15e17]. Nevertheless the mathematical tools 30 mm from the surface but was essentially concentrated in
presented here can be applied to other stones and purposes. the first 20 mm, corresponding to the grey crust [19].

2. Materials and methods 2.2. Obtaining the 2-D images

2.1. Materials We acquired 2-D images for quantitative analysis by using


scanning electron microscopy on thin sections (30  45 mm2).
In order to show how 2-D computerized image analysis of The latter were obtained after sample impregnation with
pore and solid phase distribution can be applied to monument a polyester resin under vacuum. The thin sections were pol-
conservation, quarry and weathered limestones were selected. ished and coated with carbon prior to observation using back-
Two non-weathered limestones were selected: a Tuffeau stone scattered electron emission [20]. Thin sections parallel to the
and a Sebastopol stone. stone bed were produced for the quarry stones. For the weath-
Tuffeau stone was collected in a quarry located near the ered Tuffeau stone, two thin sections were produced perpen-
village of Saint-Cyr-en-Bourg (France). Historically, Tuffeau dicular to the exposed surface: one enabling the study from
stone was chosen to build most Loire chateaux, churches, the surface to 25 mm depth and the other one, from 25 to 50 mm
cathedrals, and houses along the Loire valley. This stone is depth. Backscattered electron scanning images (BESI), con-
an easily workable building material. Today, Tuffeau stone is sisting of arrays of 736  500 pixels for quarry Tuffeau
mainly used to restore these monuments. Tuffeau stone is a stone samples (736  400 for weathered Tuffeau stone) and
yellowish-white porous sedimentary limestone mainly charac- 1024  800 pixels for Sebastopol stone samples, each pixel
terized within the last decade [15e17]. This is a siliceous having 8 bits depth (256 levels grey scale), were recorded.
limestone of Middle Turonian age (90  106 years) located The resolution was 2.8 and 3.8 mm per pixel for the Tuffeau
principally between the cities of Angers and Tours along the and Sebastopol stone samples, respectively.
Loire and the Vienne rivers. Previous studies [16,17] showed
that Tuffeau stone is essentially composed of calcite, silica 2.3. Image analysis and segmentation
in the form of opal cristobalite-tridymite and quartz and
some secondary minerals such as clays and micas. The Tuffeau One of the main problems with image analysis of porous
stone studied in this paper is composed of 50.3% calcite and materials lies in the distinction between the pore and the solid
45.2% silica (determined by induced chemical plasma) and phase. This is mainly related to the finite pixel size, noise
the total porosity obtained by density measurement was caused by data acquisition or inappropriate sample prepara-
48.1%. It is a multi-scale porous medium since the equivalent tion. For geometrical analysis of 2-D images, it is essential
pore size distribution ranged from 0.01 to 50 mm in size [18]. to apply a well defined method that permits the determination
Sebastopol stone was collected in a quarry located in the of whether a pixel belongs to the pore or to the solid phase.
north of Paris. This stone is a yellowish-beige sedimentary This procedure is known as image segmentation and must be
limestone composed essentially of calcite and quartz. It was reliable and accurate. Segmentation is a process that consists
O. Rozenbaum et al. / Journal of Cultural Heritage 8 (2007) 151e159 153

of converting a greyscale image into a binary image by iden- stone and the Sebastopol stone originating from a quarry, and
tifying two sets of pixels in the image on the basis of their grey 14 images for the weathered Tuffeau stone.
level. Usually the threshold value is chosen according to the
shape of the greyscale histogram. This procedure leads to re- 3. Results and discussion
liable results if the histograms of the two phases are clearly
separated. Homemade algorithms were implemented in Cþþ 3.1. Pore and solid phase morphology
in order to calculate the histograms, to determine the thresh-
olds and to segment the images. Twenty-nine images were Scanning electron microscopy realized on fractures at high
analysed: eight and seven images respectively for the Tuffeau resolution (not shown here) and BESI (Fig. 1a,b) showed that

Fig. 1. Backscattered electron scanning images of (a) a quarried Tuffeau stone; (b) a quarried Sebastopol stone; (c) and (d) the strongly weathered zone; (e) moderately
weathered zone of a weathered Tuffeau stone; and (f) the un-weathered zone of a weathered Tuffeau stone.
154 O. Rozenbaum et al. / Journal of Cultural Heritage 8 (2007) 151e159

the Sebastopol limestone was composed of larger grains


(mainly 1e400 mm in size) than the quarry Tuffeau stone
(mainly 0.1e150 mm). Furthermore, the Sebastopol stone
grains appeared to be less cemented than those of the Tuffeau
stone. For the weathered Tuffeau stone originating from the
church, three zones were distinguished according to both
weathering intensity and depth. BESI showed a strongly weath-
ered zone from the surface to 2 mm in depth (Fig. 1c,d) with the
presence of large pores, some of them being elongated pores
corresponding to cracks. BESI also showed a deeper and mod-
erately weathered zone from 2 mm to 30 mm in depth (Fig. 1e)
where cracks were still present, although the grains were well-
cemented compared to the strongly weathered zone. Deeper,
there were no cracks, the Tuffeau stone looked unaltered and
similar to the quarry Tuffeau stone (Fig. 1f).
Greyscale histograms recorded for the images studied were
bimodal (Fig. 2). Most pore phases corresponded to the [0e
65] greyscale range and most solid phases to the [65e255]
greyscale range. The threshold value was taken as 65, i.e. at
the minimum between the two distributions. Results showed
also that a little change in greyscale (5) for this threshold
value did not impinge strongly or cause modifications in the
resulting segmented image (Fig. 3).

3.2. Porosity

Measurement of porosity on a 2-D image requires measure-


ments on a representative elementary surface area (RESA). A
RESA is a surface area over which a statistical averaged prop-
erty can be computed. The RESA should be extracted from
Fig. 3. (a) Backscattered electron scanning image of a quarried Tuffeau stone
a large enough image size to provide representation of macro- and (b) the same image after segmentation.
scopic properties of the media. The RESA is not always caught
by experimental image acquisition and must be estimated for Classically total porosity (f) is defined as follows:
every new image. Determination of the RESA is implemented
by taking a small surface area within an image and by calcu- Vp
f¼ ð1Þ
lating the property of interest (e.g. porosity). The small surface Vp þ Vs
area is then expanded in all directions and the property is recal-
where Vp and Vs are the volume of the pore and solid phase,
culated. The RESA is determined as the surface area value
respectively. For a 2-D digitized medium, Vp and Vs are the
over which the property of interest remains constant.
numbers of pixels corresponding to the porous phase and to
the solid phase, respectively.
In order to estimate the RESA for f, porosity was measured
for different image sizes and three examples are given in
Fig. 4. As expected for the smallest area, porosity varies ex-
tremely but, as the image size increases, porosity tends to
a limit. Obviously, this limit represents porosity for pores
larger than the pixel size resolution. It can be concluded that
the RESA exists for the samples studied and is reached for
the whole image (at least for total porosity). Results showed
that the RESA corresponded to a surface area of about
2  105 pixel squares for these resolutions. This surface area
corresponded to 1.6 and 2.9 mm2 for the Tuffeau and Sebastopol
stones, respectively.
Average porosities calculated for the different stones are
reported in Table 1. The standard deviations are similar for
Fig. 2. Greyscale histogram recorded for the image of Tuffeau stone shown in the stones studied and are relatively small. The porosity of
Fig. 1e. the Sebastopol stone is slightly higher than that of the quarry
O. Rozenbaum et al. / Journal of Cultural Heritage 8 (2007) 151e159 155

Table 1
Average total porosity and standard deviation recorded for the stone samples
studied
Number Average Standard
of images porosity (%) deviation (%)
Quarried Tuffeau stone 8 30.5 2.5
Quarried Sebastopol stone 7 35.6 2.4
Weathered Tuffeau stone
Strongly weathered zone 7 45.2 2.1
Moderately weathered zone 7 28.6 2.3

pixel size resolution) and a porosity determined from density


measurement of 43.4% [18].

3.3. Autocorrelation function

The autocorrelation function is another tool enabling the


statistical description of a porous medium. In the following
we define the two-point correlation function, keeping in
mind that a correct description of correlations is done by cal-
culating the n-point correlation functions [21]. Let / x be the
/
position vector from an arbitrary origin and Jm ðx Þ be a den-
/
sity function defined as Jm ðx Þ ¼ 1 if /x belongs to the pore
/
space and Jm ðx Þ ¼ 0 if /
x belongs to the solid space. The po-
rosity f and the two-point correlation function S2 ðr/ Þ can be
defined by the statistical averages [22,23]:
/
f ¼ hJm ðx Þi ð2Þ

/ /
S2 ðr Þ ¼ hJm ðx ÞJm ðr/ þ /
x Þi: ð3Þ

The angle brackets mean surface average over the spatial co-
ordinates /x . Writing the last equation in this way assumes
that the porous medium is statistically homogeneous. In other
words, on average, only differences between two coordinate
values are important and not their absolute location. The two
limits of S2 ðr/ Þ are [24]:

S2 ð0Þ ¼ f ð4Þ

lim S2 ðr/ Þ ¼ f2 : ð5Þ


r/N

With these limits, the autocorrelation function RZ ðr/ Þ can be


defined in order to have a normalized function:
/ /
Fig. 4. Total porosity versus the surface area of the image of (a) a quarried / hðJm ðx Þ  fÞðJm ðr þ/
x Þ  fÞi
Tuffeau stone; (b) a quarried Sebastopol stone; (c) the strongly weathered RZ ðr Þ¼ : ð6Þ
fð1  fÞ
zone of a weathered Tuffeau stone.

This function can be interpreted as the probability of finding


Tuffeau stone. The porosity determined by image analysis two randomly selected points that are both in the same phase.
for the Tuffeau stone originating from the quarry (30.5%) is For an isotropic medium with r ¼ jr. j, RZ ðr/ Þ ¼ RZ ðrÞ, the au-
much smaller than the porosity determined from density mea- tocorrelation function depends on distance alone. Otherwise,
surement (48.1%) [18]. That difference is related to the small the one dimension autocorrelation functions for ! r parallel to
pores which are not taken into account in image analysis be- the Ox or Oy axis (Ox orthogonal to Oy), respectively noted
cause they are smaller than the pixel size (2.8 mm). This be- RZx , RZy give information about isotropy or anisotropy. Indeed,
haviour is also encountered with the Sebastopol stone, which pore space anisotropy is revealed as a disparity between the
has porosity determined by image analysis of 35.6% (3.8 mm one-dimensional autocorrelation functions along different
156 O. Rozenbaum et al. / Journal of Cultural Heritage 8 (2007) 151e159

directions [25]. However, the autocorrelation function does not


provide information about the connectedness of the phases.
We calculated the autocorrelation functions RZx , RZy and the
average value RZ for the different samples. An example of the
recorded autocorrelation function is given in Fig. 5. All these
functions present a decreasing behaviour without any particu-
lar correlation. Besides, a little disparity depending on the im-
ages was observed for the one-dimensional autocorrelation
functions along the orthogonal x- and y-directions. The Tuf-
feau stone (Fig. 5a) and Sebastopol stone (not shown here)
originating from the quarry both present no, or very slight, an-
isotropies. Anisotropy is greater in the apparently moderately
(Fig. 5b) and the strongly (not shown here) weathered zone of
the weathered Tuffeau stone. This can be related to the pres-
ence of cracks or to a preferential dissolution of the solid
phase that is not isotropic since it develops from the surface.
Finally, comparison between the autocorrelation functions re- Fig. 6. Autocorrelation functions of the quarried Tuffeau stone and Sebastopol
limestone.
corded for the quarry Tuffeau and Sebastopol stones confirmed
that the characteristic dimensions of the Sebastopol stone are
higher than those of the Tuffeau stone (Fig. 6).
3.4. Chord distributions

Chord distributions are stereological tools allowing the


description of the interface between pore and solid phases.
A chord is a segment belonging either to the pore or to the
solid phase and having both ends on the interface (Fig. 7).
The chord distribution gives the probability to have a chord
length between r and r þ dr. Chord distribution can be calcu-
lated either for the pores ( fp(r) is called the pore chord distri-
bution) and for the solid ( fs(r) is called the solid chord
distribution). The chords are calculated along randomly dis-
tributed lines in the porous medium. Furthermore, the first

Fig. 5. Autocorrelation functions recorded for (a) a quarried Tuffeau stone; Fig. 7. Schematic illustration of the pore chord distribution (pore phase in
(b) and the moderately weathered zone of a weathered Tuffeau stone. white and solid phase in grey).
O. Rozenbaum et al. / Journal of Cultural Heritage 8 (2007) 151e159 157

 
momentum of fp(r) (called lp) and the first momentum of fs(r) fp ðrÞhexp  r=ap ð9Þ
(called ls) are defined as:
Z and
lp ðrÞ ¼ rfp ðrÞ dr ð7Þ
fs ðrÞhexpð  r=as Þ ð10Þ
and
Z where ap is the persistence length for the porous phase and as
is the persistence length for the solid phase. These two persis-
ls ðrÞ ¼ rfs ðrÞ dr: ð8Þ
tence lengths ap and as correspond to mean distances bet-
ween two interfaces across the pore and the solid phase,
The values of lp and ls can be discussed as estimators of the respectively. Cousin and colleagues [14] suggested choosing
mean size of the pore and solid phases. an image size superior to four times the persistence length,
Results showed that the greater the chord length was, the to study the chord distributions with a representative surface
noisier the chord distribution recorded was (Fig. 8). This noise area. This type of porous medium, in which both the porous
corresponded to large phases (pore or solid) few of which were and solid chord distributions are exponential, is called
in an image of finite size. Due to digitizing, the smallest pores a long-range random medium [13]. In other words, the pore
were badly defined and did not correspond to real objects. So, and solid phases in the image were randomly distributed.
for r < 3 pixels, chord distributions presented a linear increase The absence of correlation peaks demonstrated that the pore
that had no reality and did not have to be taken into account and solid phases were heterogeneously distributed in size. In
[14]. Most pore and solid chord distributions studied presented the case of such exponential decay, ap ¼ lp and as ¼ ls.
an exponential decrease in a large range of chord lengths. The Pore and solid chord distributions of the quarry Tuffeau and
mathematical expressions of these distributions are [13]: Sebastopol stones were calculated for different images taken

Fig. 8. Pore and solid chord distribution recorded for different images of Fig. 9. Pore and solid chord distribution recorded for different images of
a quarried Tuffeau stone. a quarried Sebastopol stone.
158 O. Rozenbaum et al. / Journal of Cultural Heritage 8 (2007) 151e159

in the same thin sections for each type of stone. Distributions


presented an exponential decay (Figs. 8 and 9) but within
a larger chord length range for the Sebastopol stone. Differ-
ences between pore chord distributions for both stones were
recorded for the longest chords, related to the presence of large
pores in images. Solid chord distributions show also an expo-
nential decay except for the longest chords because of the
presence of some large grains. First moments of the chord
distributions are on average for the Tuffeau stone lpz10 mm
and ls z 22 mm and for the Sebastopol stone lp z 26 mm
and ls z 47 mm. Thus, even if total porosities were equivalent,
the pore and solid phases of the Sebastopol stone were two or
three times greater than those of the Tuffeau stone, as shown in
Fig. 10. The little variations in the whole range of the chord
length show an important point: chord distributions were
able to capture few differences between images. These devia-
tions were due to the natural heterogeneity of the geomaterials.
Nevertheless, chord distributions present similar behaviour
and are representative of the stones. Thus, chord distributions
are very sensitive to some geometrical fluctuations, making it
possible to reveal small heterogeneities in geomaterials.
Solid chord distributions enable the distinction between the
strongly weathered zone and the moderately or un-weathered
zone. Indeed, results showed ls z 14 mm for the former and
ls z 23 mm for the latter (Fig. 11). These solid chord distribu-
tions formed two sets that were easily distinguishable: solid
chord distributions of the strongly weathered zone always
have a higher persistence length (shown by slopes in semi
logarithmic representation) than the solid chord distributions
of the moderately or un-weathered zone. Thus, the distance be-
tween the two interfaces within the solid phase is statistically
greater in the moderately or un-weathered zone than in the Fig. 11. Comparison between the chord distributions of (a) the moderately
strongly weathered zone. This distinction arises from the disso- weathered zone shown in Fig. 1e; (b) the un-weathered zone shown in
lution processes, resulting in larger pores in the strongly weath- Fig. 1f; (c) and (d) the strongly weathered zones shown in Fig. 1c,d.
ered zone.
Existence of cracks does not change the solid chord distri- moderately weathered zone (Fig. 11a), where cracks were
butions but affects the pore chord distributions of the range of present, had a higher slope than the pore chord distribution
the longest chords. Indeed, pore chord distribution of the of the un-weathered zone (Fig. 11b) for chords >100 mm but
were similar for smaller chords. Up to 100 mm, the pore chord
distribution in the moderately weathered zone (Fig. 11a) could
be described by a second exponential function that was repre-
sentative of cracks. This was exactly the same remark for the
pore chord distributions of both strongly weathered zones
studied (Fig. 11c,d). Pore chord distributions of moderately
or un-weathered zones and of a strongly weathered zone
(Fig. 11a,b,d) were similar for chords < 100 mm, but for
another strongly weathered zone (Fig. 11c), pore chord distri-
bution had a lower slope than the others, showing that the dis-
tance between two interfaces within the pore phase in this
image was always statistically greater. This behaviour was
expected for moderately or un-weathered zones in comparison
to the strongly weathered zone (Figs. 1c and 11c). But, this
was surprising for the comparison between the two pore chord
distributions of strongly weathered zones. In fact, as the solid
Fig. 10. Comparison between (a) the pore chord and (b) solid chord distri- chord distributions of the two strongly weathered zones stud-
bution of a quarried Sebastopol stone; (c) the pore chord and (d) solid chord ied were similar (proving the same ‘‘rate’’ of dissolution), it
distribution of a quarried Tuffeau stone. can be assumed that presence of larges cracks (Fig. 1c)
O. Rozenbaum et al. / Journal of Cultural Heritage 8 (2007) 151e159 159

indicated a deep modification of porosity, even in the range of Natural Stone, Weathering Phenomena, Conservation Strategies and
the smallest pores (Fig. 11c). This was not the case for the Case Studies, 205, Geol. Soc. Special Publication, 2002, pp. 329e345.
[4] F. Monforti, R. Bellasio, R. Bianconi, G. Clai, G. Zanini, An evaluation
pore chord distributions of the moderately weathered zone. of particle deposition fluxes to cultural heritage sites in Florence, Italy,
Thus, thin cracks shown in the moderately weathered zone Sci. Total Environ. 334-335 (2004) 61e72.
would indicate few modifications for inter-grain pores of the [5] D.A. Dolske, Deposition of atmospheric pollutants to monuments,
Tuffeau stone. statues, and buildings, Sci. Total Environ. 167 (1995) 15e31.
Furthermore, the ratio ls/lp might also be an interesting [6] R.S. Hamilton, D.M. Revitt, K.J. Vincent, R.N. Butlin, Sulphur and nitro-
gen particulate pollutant deposition on to building surfaces, Sci. Total
criterion for the study of the effect of weathering on stones. Environ. 167 (1995) 57e66.
In images of the un-weathered zone, this ratio was on average [7] C. Rodriguez-Navarro, E. Sebastian, Role of particulate matter from
2.4 and it was 1.2 in images of the strongly weathered zone. vehicle exhaust on porous building stones (limestone) sulfation, Sci. Total
This result is quite intuitive because the mean size of the solid Environ. 187 (1996) 79e91.
phase decreases (and the mean size of the porous phase in- [8] A. Bruand, R. Prost, Effect of water content on the fabric of a soil
material: an experimental approach, J. Soil Sci. 38 (1987) 461e472.
creases) due to weathering consequences. [9] P. Renault, Theoretical studies of mercury intrusion in some networks:
testing the applicability of mercury intrusion in the size characterisation
4. Conclusion of the lacunar pore space of soil samples, Trans. Porous Media 3 (1988)
529e547.
[10] S. Brunauer, P.H. Emmett, E. Teller, Adsorption of gases in multimolec-
The results show that the mathematical tools presented in ular layers, J. Am. Chem. Soc. 60 (1938) 309e324.
this paper are able to reveal, quantitatively and statistically, [11]  Török, N. Rozgonyi, Morphology and mineralogy of weathering
A.
differences between quarried limestone. They are also able crusts on highly porous oolitic limestone, a case study from Budapest,
to clearly distinguish the degree of weathering of the stones Environ. Geol. 46 (2004) 333e349.
collected from buildings. The resulting characteristics are [12] L.P. D’Acqui, A. Bruand, M. Pagliai, Study of soil porosity with mercury
porosimetry and image analysis on backscattered electron scanning
promising information for a better understanding of weathering images (BESI). Application to tilled ‘‘crusting soils’’ in Zimbabwe.
mechanisms. Soil micromorphology: studies in management and genesis. Proceedings
Chord distributions enable the study of pore and solid phase of the IX International Working Meeting on Soil Micromorphology,
repartition regardless of total porosity and thus show very Townsville, Australia, July 1992, Dev. Soil Sci. 22, pp. 581e590.
small differences in pore morphology and pore structure be- [13] P. Levitz, D. Tchoubar, Disordered porous solids: from chord distribution
to small angle scattering, J. Phys I. 2 (1992) 771e790.
tween quarry stones or different zones of a weathered stone. [14] I. Cousin, P. Levitz, A. Bruand, Three-dimensional analysis of a loamy-
Autocorrelation function analysis showed that quarried stones clay soil using pore and solid chord distributions, Eur. J. Soil Sci. 47
appeared as being an isotropic porous geomaterial at the study (1996) 439e452.
scale for 2-D images taken parallel to stone bed. Discussion of [15] K. Beck, M. Al-Mukhtar, O. Rozenbaum, M. Rautureau, Characteriza-
isotropy in the direction perpendicular to stone bed will re- tion, water transfer properties and deterioration in Tuffeau stone: build-
ing material in the Loire valley, France, Building Environ. 38 (2003)
quires 2-D images taken in that perpendicular direction. How- 1151e1162.
ever, discussion of porosity characteristics with the aim of [16] D. Dessandier, Etude du milieu poreux et des propriétés de transfert des
better understanding of the weathering mechanisms and, fluides du Tuffeau blanc de Touraine. Application à la durabilité des
more generally, the transfer properties of the stone being pierres en œuvre, PhD thesis, University of Tours, France, 1995.
studied, requires 3-D analysis of the porosity characteristics. [17] B. Brunet-Imbault, Etude des patines de pierres calcaires mises en oeuvre
en Région Centre, PhD thesis, University of Orléans, France, 1999.
Mathematical tools applied to 2-D images in this study can [18] A. Arrizi, Personal communication, 2005.
be easily extended to 3-D images. This will be the next step: [19] S. Anne, Analyse de l’altération d’un Tuffeau stone mis en œuvre,
using 3-D images obtained by X ray microtomography of the Master of Science and Technology, University of Orléans, France, 2005.
same stone samples. [20] A. Bruand, O. Duval, Calcified fungal filaments in the petrocalcic
horizon of Eutrochrepts in Beauce (France), Soil Sci. Soc. Am. J. 63
(1999) 164e169.
References [21] S. Torquato, Random Heterogeneous Materials, Springer, 2002.
[22] J.G. Berryman, Measurement of spatial correlation functions using image
[1] D. Camuffo, Physical weathering of stones, Sci. Total Environ. 167 processing techniques, J. Appl. Phys. 57 (1985) 2374e2384.
(1995) 1e14. [23] J.G. Berryman, S.C. Blair, Use of digital image analysis to estimate fluid
 Török, Oolitic limestone in a polluted atmospheric environment in
[2] A. permeability of porous materials: Application of two-point correlation
Budapest: weathering phenomena and alterations in physical properties, functions, J. Appl. Phys. 60 (1986) 1930e1938.
from natural stone, weathering phenomena, conservation strategies and [24] D.A. Coker, S. Torquato, Extraction of morphological quantities from a
case studies, Geol. Soc. London. 205 (2002) 363e379. digitized medium, J. Appl. Phys. 77 (1995) 6087e6099.
[3] R.A. Lefèvre, P. Ausset, Atmospheric pollution and building materials: [25] P. Frykman, T.A. Rogon, Anisotropy in pore networks analyzed with 2-D
stone and glass, in: S. Siegesmund, T.Y. Weiss, A. Vollbrect (Eds.), autocorrelation (variomaps), Comput. Geosci. 19 (1993) 887e930.