GIScience Integrated with Computer Vision for the Interpretation

and Analysis of Old Paintings

Motti Zohar

, Ilan Shimshoni

and Fadi Khateb

Department of Geography and Environmental Studies, University of Haifa, Israel

Department of Information Systems, University of Haifa, Israel

Keywords: GIScience, Computer Vision, Geo-referencing, Cartography, Jerusalem.

Abstract: Photographs of Ottoman Palestine are available only from the 2

half of the 19

century onward. Thus, in

order to reconstruct the landscape at the time one should rely on other visual sources such as old paintings.

To do so, their accuracy and completeness must be addressed first. In this paper we analyse a painting from

1823 by the British Sir Frederick Henniker that drew the Old City of Jerusalem when standing somewhere on

the Mount of Olives. We use GIScience techniques with computer vision capabilities to resolve the exact

location where the artist stood as well as verifying errors and completeness of the painting. Preliminary results

demonstrate that the location of the artist when drawing the painting was on top of Mount of Olives (close to

present-day 7 arches hotel) rather at the Cave of the Apostles as cited in the NLI librarian citation. Additionally,

the accuracy of his work was verified by comparing the features he drew on the canvas to their actual location

on a present-day photograph.

1 INTRODUCTION

In 1820 the British traveller, Sir Frederick Henniker

(1793-1825), visited Ottoman Palestine (Ben-Arieh,

1997, p. 30). In his book from 1823 he presented a

painting of the Old City of Jerusalem alleged to be

drawn from the Cave of the Apostles located at the

western slope of the Mount of Olives (Henniker,

1823). Henniker portrayed the Old City surrounded

by the 16

century Ottoman walls along with other

prominent structures and monuments, some are still

standing till present-day (Fig. 1). Photography in

Palestine was not available until the 2

half of the 19

century (Nir, 1985). Thus, Henniker’s painting as

well as paintings of other artists (Röhricht, 1890) may

serve as a meaningful resource for past landscape

reconstructions in periods prior to the mid-19

century. However, relying on historical paintings

must be done carefully whereas they might be

inaccurate, incomplete and subjected to the artist’s

perception and conceptualization of reality.

https://orcid.org/0000-0002-6567-7296

https://orcid.org/0000-0002-5276-0242

Figure 1: (a) The Old City of Jerusalem photographed from

the Mount of Olives (photographing: Motti Zohar, January

2020). Prominent features are noted in black on the

photograph e.g., E (Al Aqsa mosque), L (Dome of the

Rock), M (Church of the Holy Sepulchre) and Q (the

Custodia Terra Sancta); (b) The painting of Henniker from

1823. See the complete list of notations in Table 1 in the

Appendix.

Zohar, M., Shimshoni, I. and Khateb, F.

GIScience Integrated with Computer Vision for the Interpretation and Analysis of Old Paintings.

DOI: 10.5220/0009464902330239

In Proceedings of the 6th International Conference on Geographical Information Systems Theory, Applications and Management (GISTAM 2020), pages 233-239

ISBN: 978-989-758-425-1

233

2 GIScience AND COMPUTER

VISION

The rapid development of GIS (Geographic

Information Systems) and GIScience (Geographic

Information Science) during the last few decades

have revolutionized the way scientists manage and

analyse spatial data (Goodchild, 2010). With

quantitative and qualitative capabilities (Cope &

Elwood, 2009; Y. Wang & Taylor, 2018) GIS serves

a wide range of spatial applications (Sui, 2015) and

forms a proper platform to verify spatial hypotheses.

As far as cartography is concerned, GIScience

enables spatial analyses (Levin, 2006; Schaffer et al.,

2016; Zohar, 2019); the creation of deep maps as

multimedia conveyors of places and their everyday

lives (Bodenhamer et al., 2015); and the inspection of

narratives using map stories (Antoniou et al., 2018;

Mennis et al., 2013). In historical-based studies,

GIScience is used in creating new forms of virtual

knowledge (Gregory & Healy, 2007; Knowles,

2008); performing 2D and 3D landscape

reconstructions (e.g., Davie & Frumin, 2007;

Georgoula et al., 2013; Nakaya et al., 2010;

Rubinowicz & Czyńska, 2015; Zohar, 2017); and

resolving complex scenarios of past phenomena (e.g.,

Bender et al., 2005; Katz & Crouvi, 2007; Verhagen

& Jeneson, 2012).

Like GIS, computer vision and especially

machine learning based computer vision have

undergone major advances over the last two decades

enriching also the GIS capabilities and automating

tedious processes such as feature extraction and

digitization (e.g., Chen et al., 2018; Fanos et al., 2018;

Garosi et al., 2019; Naghibi et al., 2016). Methods can

therefore be applied and developed for addressing

problems of interpretation of historical visual sources.

These include methods for registration of

photographs as an addition to geo-referencing maps

(Aubry et al., 2014; Goshen & Shimshoni, 2008;

Hartley & Zisserman, 2003). Once a given source has

been aligned and registered, the use of machine

learning may assist in extracting and aligning features

such as roads and buildings (Duan et al., 2017; Uhl et

al., 2017; J. Wang et al., 2015).

The goal of the study is to interpret the old

painting of Henniker using GIScience approaches

integrated with computation vision capabilities. In

this short paper we attempt to address the following

two questions: (1) What used to be the location of the

artist when he drew the painting? (2) What are the

errors made by the artist in comparison to actual view

of Jerusalem? We hypothesize that computer vision

capabilities may empower and semi-automate the

GIScience approaches initiated so far in analysing old

paintings during the processes of geo-referencing,

digitization and errors verification. Accordingly, we

present preliminary results of the implemented

methodology aimed at addressing these problems of

interpreting old paintings.

3 METHODOLOGY

3.1 The GIS Environment

We use the ESRI

ArcGIS Pro software as our base

platform for the analysis and the Israel Transverse

Mercator (ITM) (Mugnier, 2000) as the preferred

Coordinate Reference System (CRS). For coding we

use Python which can be run both internally within

the ArcGIS Pro suit and by an external IDE

(Integrated Development Environment). The

geospatial data we have used as GIS layers for our

analysis is (1) present-day orthophoto of Jerusalem

with a resolution of 0.35 meter/pixel; (2) The mid-19

century Jerusalem map of Wilson (1865); and (3)

Digital Elevation Model (DEM) of the Advanced

Land Observation Satellite–Phased Array type L-

band Synthetic Aperture Radar (ALOS-PALSAR)

with a resolution of 12.5 meter/pixel

(http://www.eorc.jaxa.jp/ALOS/en/about/palsar.htm)

downloaded from https://vertex.daac.asf.alaska.edu/#

3.2 3D Features as Control Points

We have noted as control points 22 prominent

features appearing in the painting of Henniker as well

as in the map of Wilson and still exist till present-day

(Figures 1, 2 and Table 1 in the Appendix). For each

feature, we have extracted the parameters of

longitude, latitude and absolute elevation (above sea

level) using the DEM. Additionally, we have

calculated the height above surface of each feature

using measurements taken in field and also those

made by Alud and Hillenbrand (2000). Given the

orientation of the painting as looking from the Mount

of Olives from east to west towards the Old City of

Jerusalem (Fig. 1), a study area of nearly 1 square

kilometre, elongated south to north, was delineated

along the ridge of Mount of Olives (Fig 2) as a

plausible area in which the artist might have stood

while he was completing his work.

3.3 Line of Sight

We assume the artist was painting what he was able

to observe. In other words, the 22 noted features were

GISTAM 2020 - 6th International Conference on Geographical Information Systems Theory, Applications and Management

234

probably visible to him from the location of drawing.

Accordingly, the delineated study area was gridded

into cells of 0.25*0.25 meters, resulting in 158412

potential observation points along the Mount of

Olives. Then, a visibility line of sight for each of the

potential observation point was evaluated. That is,

how many of the 22 noted features can be seen from

each of the observation points. The visibility

evaluation took into consideration a 6-8 m height of

the surrounding Ottoman walls as sight obstacles and

an offset of 2 meters above the surface as the observer

height. The frequency of the visible features from the

observation points that are within the delineated study

area is presented in Figure 3. Accordingly, the area in

which all 22 features are visible is portrayed in black

while region with limited visibility are portrayed in

purple. Not surprisingly, the visibility results are

primarily dictated by height differences, although

artificial structures such as buildings may also

influence but were omitted from the analysis.

Figure 2: Aerial photography of present-day Old City of

Jerusalem and the delineated study area on the Mount of

Olives (outlined in red). The alleged location of the artist

(Cave of the Apostles) as cited in the librarian record of the

NLI (National Library of Israel) is presented a purple

square.

3.4 Computer Vision Feature Matching

Assuming the painting is equivalent to a photograph,

a robust RANSAC (Fischler & Bolles, 1981; Hartley

& Zisserman, 2003) algorithm was implemented,

which recovered the position, orientation, and

parameters of the "camera" (the artist position). In

theory it is impossible to separate the focal length of

the camera (it’s zoom) and the distance from the

scene. Therefore, constraints on the artist’s position

were given in association with a given observation

point. This is important for verifying the horizontal

and vertical accuracy of the painting.

Figure 3: Visibility of the 22 noted features in and around

the Old City of Jerusalem. The potential observation points

where all features are visible are depicted in black.

4 PRELIMINARY RESULTS

4.1 The Location of the Artist

According to the NLI (National Library of Israel)

citation, the alleged location where the artist stood

when he drew the painting was at the Cave of the

Apostles on the western slope of Mount of Olives (at

elevation of 720 m above sea level) (Figure 2).

Following the iteration of the RANSAC model on

each of the potential observation points, it was found

that the observation point with the highest score is on

top of the mountain at coordinates 631574N,

223120E (at elevation of 794 m above sea level, close

to present-day Seven Arches Hotel) (Figure 3). The

elevation difference of 74 meters between the alleged

and calculated locations explains why the Cave of the

Apostles is not a reasonable location; had the artist

stood at this position, he could have barely seen any

of the features within the Old City. In other words, the

relative height inferiority of the Cave of the Apostles

in compare to the Old City and the surrounding

GIScience Integrated with Computer Vision for the Interpretation and Analysis of Old Paintings

235

Ottoman walls, blocks the possibility to observe any

structure located within the city. On the other hand,

our calculated location (using the RANSAC model)

confirms a clear line of sight from this observation

point to each of the 22 features as demonstrated in

Figure 1a that was photographed from the same

location.

4.2 Errors Made by the Artist

Figure 4a presents the identification of the 22 features

noted on the painting and the equivalent identification

of the RANSAC model (green and red lines), which

corresponds to the observation point with the highest

suitability score (Figure 3). For comparison, we have

implemented a RANSAC model on a present-day

photograph (Figure 4b) taken from the same position.

The results of the RANSAC models on both the

painting and photograph are used for orientation and

identifying notation errors (on the photograph) and

errors made by the artist (in the painting).

Table 1 (see Appendix) presents the differences in

pixels between the feature notations in compare to the

notations achieved by the RANSAC model. The

errors (in pixels) are listed in fields EH and EP for

Henniker’s painting and the photograph with average

error values of 56 and 51 pixels, respectively. The

errors distributions were then classified into quartiles

and the those included in 4th quartile were inspected.

Accordingly, the features with higher error values are

T, U and of lesser extent G and O on Henniker’s

painting and F and R on the present-day photograph

(Table 1 in the Appendix). The prominent high

elevated feature Q (the Custodia Terra Sancta) on the

present-day photograph was not detected on

Henniker’s’ painting; perhaps it may explain an error

of the artist when portraying a none-identified

minaret at the northwestern corner of the Old City.

The errors of features T, U and O are 195, 196 and 89

pixels, respectively. These features are located at the

north of the city (on the right wing of the painting)

imply that the artist may have overestimated lengths

at the northern region which offsets his vision.

Clearly the center of the city was more visible to the

artist thus enabling better approximation of lengths

from the location he was standing when drawing the

painting (Figure 3). The slight errors of the RANSAC

model on the present-day photograph may result from

potential distortion of the focal lens and will be

verified during the next stages of the study. Overall,

the average error of the Henniker painting is slightly

bigger than that of the photograph (56 pixels in

comparison to 51 pixels) implying that the artist was

quite accurate when drawing the scene and the

included prominent features.

5 CONCLUSIONS

In this paper we present preliminary results of a

methodology combining GIScience and computer

vision capabilities to cope with the inaccuracy

associated with historical visual sources such as old

paintings. The methodology used is at its first stages

and is being currently further developed. Yet, it

enables already at this point to track the position of

the artist when he drew the painting thus

contradicting the alleged location to be at the Cave of

the Apostles as cited in the painting’s library record

at the NLI. We have also managed to verify the

accuracy of features on the painting’s canvas, perhaps

drew by the artist with less accuracy due to sharp

offset from the location he was standing or in order to

meet his subjective perception. Additionally, we were

able to identify errors on the painting in compare to

present-day photograph by comparing the two sets of

results. Yet, to this point there are some vague

identification in Henniker’s painting which we could

have not resolved by the RANSAC model such as the

Custodia Terra Sancta location and the anonymous

minaret at the north section of the Old City.

The potential contribution of establishing

complete framework for analysing old paintings and

illustrations is enormous for historical-based studies.

In the next stages we will test additional paintings and

attempt to spatially register the included features

within a known Coordinate Reference System (CRS).

We anticipate this process will re-project the features

on the canvas of the painting back to the GIS

framework. We will test our results with past and

present-day photographs.

GISTAM 2020 - 6th International Conference on Geographical Information Systems Theory, Applications and Management

236

Figure 4: Noted features on Henniker’s painting (a) and present-day photograph (b) The lines represent the RANSAC

identification whereas green lines denote small-scale error detection while red lines stand for relatively large errors. For

complete results see also Table 1 in the Appendix.

REFERENCES

Alud, S., & Hillenbrand, R. (2000). Ottoman Jerusalem

(Vol. 1). London: Altajir World of Islam Trust.

Antoniou, V., Ragia, L., Nomikou, P., Bardouli, P.,

Lampridou, D., Ioannou, T., Kalisperakis, I., &

Stentoumis, C. (2018). Creating a story map using

geographic information systems to explore

geomorphology and history of Methana peninsula.

ISPRS International Journal of Geo-Information,

7(12), 484.

Aubry, M., Maturana, D., Efros, A. A., Russell, B. C., &

Sivic, J. (2014). Seeing 3d chairs: exemplar part-based

2d-3d alignment using a large dataset of cad models.

Ben-Arieh, Y. (1997). Painting the Holy Land in the

Nineteenth Century. Jerusalem: Yad Izhak Ben-Zvi

Publications.

Bender, O., Boehmer, H. J., Jens, D., & Schumacher, K. P.

(2005). Using GIS to analyse long-term cultural

landscape change in Southern Germany. Landscape

and Urban Planning, 70(1–2), 111-125. doi:

10.1016/j.landurbplan.2003.10.008.

Bodenhamer, D. J., Corrigan, J., & Harris, T. M. (2015).

Deep maps and spatial narratives: Indiana University

Press.

Chen, W., Li, H., Hou, E., Wang, S., Wang, G., Panahi, M.,

Li, T., Peng, T., Guo, C., & Niu, C. (2018). GIS-based

groundwater potential analysis using novel ensemble

weights-of-evidence with logistic regression and

functional tree models. Science of the Total

Environment, 634, 853-867.

Cope, M., & Elwood, S. (2009). Qualitative GIS: a mixed

methods approach: Sage.

Davie, F. M., & Frumin, M. (2007). Late 18th century

Russian Navy maps and the first 3D visualization of the

walled city of Beirut. e-Perimetron, 2(2), 52-65.

Duan, W., Chiang, Y.-Y., Knoblock, C. A., Jain, V.,

Feldman, D., Uhl, J. H., & Leyk, S. (2017). Automatic

alignment of geographic features in contemporary

vector data and historical maps.

Fanos, A. M., Pradhan, B., Mansor, S., Yusoff, Z. M., & bin

Abdullah, A. F. (2018). A hybrid model using machine

learning methods and GIS for potential rockfall source

identification from airborne laser scanning data.

Landslides, 15(9), 1833-1850.

Fischler, M. A., & Bolles, R. C. (1981). Random sample

consensus: a paradigm for model fitting with

applications to image analysis and automated

cartography. Communications of the ACM, 24(6), 381-

395.

Garosi, Y., Sheklabadi, M., Conoscenti, C., Pourghasemi,

H. R., & Van Oost, K. (2019). Assessing the

performance of GIS-based machine learning models

with different accuracy measures for determining

GIScience Integrated with Computer Vision for the Interpretation and Analysis of Old Paintings

237

susceptibility to gully erosion. Science of the Total

Environment, 664, 1117-1132.

Georgoula, O., Stamnas, A., Patias, P., Georgiadis, C., &

Fragkoulidou, V. (2013). Historical coastal urban

landscapes digital documentation and temporal study

with 2D/3D modeling functionality: The case of

Thessaloniki, Greece. Journal of Cultural Heritage,

14(5), 396-402.

Goodchild, M. F. (2010). Twenty years of progress:

GIScience in 2010. Journal of spatial information

science, 2010(1), 3-20.

Goshen, L., & Shimshoni, I. (2008). Balanced exploration

and exploitation model search for efficient epipolar

geometry estimation. IEEE Transactions on Pattern

Analysis and Machine Intelligence, 30(7), 1230-1242.

Gregory, I. N., & Healy, G. R. (2007). Historical GIS:

structuring, mapping and analysing geographies of the

past. Progress in Human Geography, 31(5), 638-653.

doi: 10.1177/0309132507081495.

Hartley, R., & Zisserman, A. (2003). Multiple view

geometry in computer vision: Cambridge university

press.

Henniker, F. (Cartographer). (1823). Jerusalem from the

Cave of the Apostles on the Mount of Olives. Drawn by

F. Henniker. Engraved by G. Hunt.

Katz, O., & Crouvi, O. (2007). The geotechnical effects of

long human habitation (2000 < years): Earthquake

induced landslide hazard in the city of Zefat, northern

Israel. Engineering Geology, 95(3-4), 57-78.

Knowles, A. K. (2008). Placing History: How Maps,

Spatial Data, and GIS are Changing Historical

Scholarship. Redlands, California: ESRI Press.

Levin, N. (2006). The Palestine exploration fund map

(1871–1877) of the holy land as a tool for analyzing

landscape changes: the coastal dunes of Israel as a case

study. The Cartographic Journal, 43(1), 45-67. doi:

10.1179/000870406X93508.

Mennis, J., Mason, M. J., & Cao, Y. (2013). Qualitative GIS

and the visualization of narrative activity space data.

International Journal of Geographical Information

Science, 27(2), 267-291.

Mugnier, C. J. (2000). Grids and datums: the state of Israel.

Photogrammetric Engineering & Remote Sensing,

66(8), 915-917.

Naghibi, S. A., Pourghasemi, H. R., & Dixon, B. (2016).

GIS-based groundwater potential mapping using

boosted regression tree, classification and regression

tree, and random forest machine learning models in

Iran. Environmental monitoring and assessment,

188(1), 44.

Nakaya, T., Yano, K., Isoda, Y., Kawasumi, T., Takase, Y.,

Kirimura, T., Tsukamoto, A., Matsumoto, A., Seto, T.,

& Iizuka, T. (2010). Virtual kyoto project: Digital

diorama of the past, present, and future of the historical

city of Kyoto Culture and computing (pp. 173-187):

Springer.

Nir, Y. (1985). The beginings of photography in the Holy

Land. Cathedra, 38, 67-80.

Röhricht, R. (1890). Bibliotheca Geographica Palaestinae:

Chronologisches Verzeichnis der von 333 bis 1878

Verfassten Literatur über das Heilige Land. London:

John Trotter Publishers.

Rubinowicz, P., & Czyńska, K. (2015). Study of city

landscape heritage using Lidar data and 3D-city

models. Paper presented at the 36th International

Symposium on Remote Sensing of Environment Berlin,

Germany.

Schaffer, G., Peer, M., & Levin, N. (2016). Quantifying the

completeness of and correspondence between two

historical maps: a case study from nineteenth-century

Palestine. Cartography and Geographic Information

Science, 43(2), 154-175. doi:

10.1080/15230406.2015.1029519.

Sui, D. (2015). Emerging GIS themes and the six senses of

the new mind: is GIS becoming a liberation

technology? Annals of GIS, 21(1), 1-13. doi:

10.1080/19475683.2014.992958.

Uhl, J. H., Leyk, S., Chiang, Y.-Y., Duan, W., & Knoblock,

C. A. (2017). Extracting human settlement footprint

from historical topographic map series using context-

based machine learning.

Verhagen, J., & Jeneson, C. F. (2012). A Roman Puzzle.

Trying to find the Via Belgica with GIS.

Wang, J., Song, J., Chen, M., & Yang, Z. (2015). Road

network extraction: A neural-dynamic framework

based on deep learning and a finite state machine.

International Journal of Remote Sensing, 36(12), 3144-

3169.

Wang, Y., & Taylor, J. E. (2018). Coupling sentiment and

human mobility in natural disasters: a Twitter-based

study of the 2014 South Napa Earthquake. Natural

hazards, 92(2), 907-925.

Wilson, C. W. S. (Cartographer). (1865). Ordnance Survey

of Jerusalem; By Captain W. Wilson, R.E. under the

direction of Colonel Sir Henry James...1864-5.

Engraved at the Ordnance Survey Office, Southampton

under the direction of Colonel J. Cameron, R.E.

Zohar, M. (2017). A city hit by an earthquake: an HGIS

approach to reconstructing the damage in Tiberias

(Israel) in 1837. International Journal of Geographical

Information Science, 31(1), 81-99. doi:

10.1080/13658816.2016.1188933

Zohar, M. (2019). Follow the road: historical GIS for

evaluating the development of routes in the Negev

region during the twentieth century. Cartography and

Geographic Information Science, 46(6), 532-546. doi:

10.1080/15230406.2019.1577176.

GISTAM 2020 - 6th International Conference on Geographical Information Systems Theory, Applications and Management

238

APPENDIX

Table 1: RANSAC model results for identifying features appearing in the painting of Henniker. Fields of the table: S - feature

notation; Name – the name of the feature; H – the height of the feature above surface; X, Y, Z - longitude, latitude and

elevation above sea level of the features, respectively; pXH, pYH - pixel coordinates of human notation on Henniker’s

paintings; rXH, rYH - pixel coordinates of the RANSAC model on Henniker’s painting; EH – pixel length difference

between human and model notation. The biggest differences of features T, U and of lesser extent G and O are noted in bold;

pXP, pYP - pixel coordinates of human notation on the photograph; rXP, rYP - pixel coordinates of the RANSAC model

on the photograph; EP – pixel length difference between human and model notation. The biggest differences of features F

and R are noted in bold. Summary values of A (average), M (Median), Mi (Min), Mx (Max) and 1Q, 2Q, 3Q (quartiles

thresholds) are also listed.

S Name H X Y Z pXH pYH rXH rYH EH pXP pYP rXP rYP EP

A Al‐NabiDau'd 12 221795 631010 795 283 213 288 159 54 

B Wall 10 222186 631291 751 637 341 582 288 76 236 1580 88 1518 16

C OmarCaliph 15 221987 631380 787 754 227 728 200 38 832 1496 811 1450 51

D Al‐Qalaa' 15 221668 631431 793 832 173 809 203 38 1091 1463 1083 1445 19

E Al‐Aqsa 15 222414 631468 752 857 287 824 311 4 1007 1582 1000 1586 8

F Davidcitadel 15 221666 631525 796 929 183 896 203 38 1198 1478 1352 1454 156

G Wall 10 222100 631198 753 396 315 473 271 89 

H Al‐Faqriah 15 222315 631620 764 1043 239 1064 278 44 1794 1554 1832 1559 38

I Al‐Omar 18 221835 631681 786 1056 198 1050 227 29 1819 1499 1810 1490 13

J GoldenGate 8 222559 631809 750 1450 363 1455 354 2 3086 1722 3113 1684 47

K Wall 10 222594 631489 726 858 376 859 425 49 1022 1771 1020 1788 17

L HarmAl‐Sharif 20 222382 631702 768 1158 226 1215 276 76 2201 1520 2252 1573 74

M HolySepulchre 20 221839 631752 792 1105 207 1120 219 19 1978 1500 1996 1491 20

N BabAl‐Asbat 12 222482 631943 756 1661 307 1585 332 8 3519 1618 3425 1655 12

O Wall 10 222511 632334 772 2166 389 2124 312 88 

P Al‐Malla'wiyah 8 222194 631960 778 1433 261 1437 263 4 2955 1592 2878 1562 82

 CustodiaT.S. 20 221620 631859 807 1145 195 2053 1423 2047 1471 49

R Lion'sGate 8 222538 632020 756 1795 379 1742 342 64 4010 1717 3842 1677 174

S SheichReichan 10 222189 632065 786 1608 237 1558 249 52 

T Wall 10 222036 632200 794 1784 255 1589 239 195 3154 1528 3198 1547 48

U Herod'sGate 8 222224 632253 785 1942 321 1755 262 196 

V SantaAnna 12 222495 632081 762 1795 341 1788 324 19 

    

    A 56 A 51

    M 44 M 43

    Mi 2 Mi 8

    1Q 19 1Q 17

    2Q 41 2Q 43

    3Q 67 3Q 57

    Mx 196 Mx 174

GIScience Integrated with Computer Vision for the Interpretation and Analysis of Old Paintings

239