International Journal of Contemporary Architecture ”The New ARCH“ Vol. 5, No.

1 (2018) ISSN 2198-7688

___________________________________________________________________________________________________________

DOI: 10.14621/tna.20180105

Developing 3D Box Counting Approach for Built Form Visual Assessment

Ehab Momin Mohammed Sanad1, Seif El din Sadig Hassan2
1)
College of Architecture and Planning, Qassim University
Saudi Arabia; [email protected]
2)
Faculty of Architecture, University of Khartoum
Sudan; [email protected]

Abstract 1. Introduction

This research aims to develop a computational approach for Quantitative visual assessment approaches acquired
visual assessment of built form reducing interventional increasing importance in the fields of town planning
subjectivity. Such an approach is supposed to serve as an legislations, heritage preservation and architectural
accurate practical tool of built form assessment for different design. Several tools and techniques emerged through
purposes, such as legal action and identity preservation.
the last three decades. It appears that fractal geometry
Fractal geometry has dominated the field of visual assessment
recently as a computational tool; fractal dimension is the
as a computational tool has dominated the field of
mathematical representation of visual complexity of quantitative assessment recently. Fractal geometry has
architecture. Only two-dimensional box counting approaches been widely used to calculate the fractal dimension for
have been applied and tested in both manual and 2D facades, which is the mathematical representation of
computerized applications. In this paper, a 3D box counting visual complexity.
approach has been developed and manually applied on Robie
house, to calculate fractal dimension, comparing it to the Fractal dimension is a quantitative technique of visual
previous two-dimensional results for the same building. The analysis that has been widely used in different fields
results found mathematically and architecturally valid. The including architecture; Bovill, C. used it as a tool to
developed approach needs to computerization in future for calculate visual complexity of buildings' façades via two-
more accuracy. dimensional images [2]. The calculation of fractal
dimension can be performed using the box counting
method. Tucker, C. [11] used an automated technique
developed from the work of Bovill, C. [2], using a
methods developed by Foroutan-pour, K. [3] to
determine the best approximation for the fractal
dimension. Ostwald and Vaughan fine-tuned and
refined the 2D box counting approach, reducing its
errors, through the application of comparative analysis
[12] retesting Bovill’s method [13] and optimizing pre-
processing standards [9].

2. Objectives

The objectives of this research are about developing the

visual domain in architecture by investigating built form,
through the development of a method of visual
investigation, which can be used in both practice and
Keywords: Built form; Computational visual
assessment; Fractal dimension; Box research.
counting The major objective of this research is to generate an
approach to investigate the visual appearance of built
Article history: Received: 02 February 2018 form, reducing interventional subjectivity within
Revised: 21 February 2018
architectural visual judgment. This approach should
Accepted: 13 March 2018
achieve the following:

___________________________________________________________________________________________________________

F. Zsuzsanna: “Local Database as Crucial Factor of Sustainable Architecture”, pp. 39–49 39

International Journal of Contemporary Architecture ”The New ARCH“ Vol. 5, No. 1 (2018) ISSN 2198-7688
___________________________________________________________________________________________________________

• To employ computational tools that based on The idea of fractal dimension emerged when
the nature of architectural form, in order to Mandelbrot, B. [7], presented the measured dimension
generate quantitative representation of visual through a discussion of the length of the coast of
quality. England. According to Li, J. [4], Fractal dimension (FD) is
a useful feature for texture segmentation, shape
• To enable visual comparison benchmarking by
classification, and graphic analysis in many fields. The
being mathematically and architecturally valid.
box-counting approach is one of the frequently used
• To provide a guide for building and planning techniques to estimate the fractal dimension of an
legislations, this means that it has to be image. A fractal dimension is a ratio providing a
practical and accurate. statistical index of complexity comparing how detail in a
fractal pattern changes with the scale at which it is
• To involve digital tools in visual form
measured.
assessment of architecture by making the
approach programmable as an algorithm.
3.3. Concept of magnification
3. Fractal geometry Geometrical topological dimension represents how
many dimensions the figure contains, but
Fractal geometry is a term founded in the late 1970's by
mathematically, a one dimensional object if magnified 2
Mandelbrot, B. [6]. He stated that "a fractal is a rough
times, will result 2 times the original (Figure1-a). In
fragmented geometric figure that can be split into parts,
addition, a two dimensional object magnified by two will
each of which is (at least approximately) a reduced-size
result 4 times the original (Figure1-b), and a three
copy of the whole; and it has two properties: self-
dimensional object magnified 2 times will result 8 times
similarity and fractal dimensions" [6] or in other words,
the original (Figure1-c). The magnification is constant,
their topological dimension is not an integer. He coined
but the result changes in accordance with the
the term fractal from the Latin adjective "fractus".
topological dimension. This means that the topological
Fractals provided a mathematical model for many
complex objects found in nature such as coastlines,
mountains, and clouds. Mandelbrot, B. [6], discussed
that Euclidian geometry, is unable to describe nature
complexity, because mountains are not conical, clouds
are not spherical, and rivers are not orthogonal; these
objects are too complex to possess characteristic sizes
and to be described by traditional Euclidean geometry Figrue 1-a. Dimensional line segment
[4]. with a magnification of 2 [15]

3.1. Fractal visual analysis of architecture

According to Ostwald, M. [8], fractal analysis is one of

the most widely used techniques for quantitative visual
assessment for the built environment character. Many Figure 1-b. Dimensional square and triangle
scholars have applied fractal geometry to analyze non- with magnification of 4 [15]
linear urban forms at the regional, city, and building
levels [5]. According to Tucker, C. [10], architectural
designers adopted fractal geometry shortly after its
emergence by Mandelbrot, B. [6]. Later it has been
involved in built environment visual analysis by many
authors. According to Ostwald, M. [8], "since 1990s,
fractal geometry has been used to measure the
properties of town plans and skylines" [8]. Architectural
researchers have also used manual fractal analysis with
different manipulations, to measure the visual
properties and visual complexity of contemporary
architecture and historic buildings.

3.2. Fractal dimension Figure 1-c. Dimensional cube

with the magnification of 2 [15]
___________________________________________________________________________________________________________

F. Zsuzsanna: “Local Database as Crucial Factor of Sustainable Architecture”, pp. 39–49 40

International Journal of Contemporary Architecture ”The New ARCH“ Vol. 5, No. 1 (2018) ISSN 2198-7688
___________________________________________________________________________________________________________

dimension also represents how a figure would scale up. Bovill's process: The process of calculating the Box-
The mathematical analysis shows form the previous counting dimension of the Robie house was performed
values that the resulting shape multiplications is equal in the following steps:
to the magnification factor raised to the power of the
Superimpose a grid of square boxes over the image (grid
dimension value.
size s1)
From the previous examples, it is clear that the
Count the number of boxes that contains some of the
mathematical relationship between "D" as the
image (N(s1))
dimension, "N" as the number of multiplications, and
"e" as the magnification factor, is according to the Repeat this procedure with smaller and smaller grid size
following equation [14]: (s2,3,and 4)
Then apply the equation:
eD = N (1)
Db == [log (N(s2)) – log (N(s1))]
D (4)
Solving the equation for D, the formula becomes [14]: b
[log (s2) – log (s1)]
D = log N / Log e (2)
For the Robie house, Figure 2 illustrates the procedure
of box-counting as performed by Bovill, [2]. The fractal
But applying the equation for a fractal shape such as the
dimension of the Robie house elevation is about 1.44 or
Kock curve will result in a fractional dimension. Because
a bit more.
with a magnification factor 3, the resulting
multiplications of the details are 4. Applying the
equation: 4. Title

D = log 4 / log 3 = 1.26 (3) The third dimension needs to be involved to complete
the visual field and to generate results that can be valid
This is why they call it a fractal dimension. This equation as a simulation of our three dimensional world. The
is known as Hausdorff-Besicovitch dimension equation human eye can receive and interpret a three
to calculate the self-similarity dimension Ds [14]. dimensional image. If the visual assessment approach is
about measuring the visual impact of the building, and
to simulate the perceptual algorithms of the human eye;
3.4. Bovill's method to calculate fractal then the third dimension has to be involved. Because a
dimension for architecture two-dimensional visual assessment may not generate
perfect results that can represent our visual interaction.
Bovill, C. [2] performed box counting method for To involve the third dimension, a complete three
calculating fractal dimension, to determine the dimensional visual assessment is the best way, that can
approximate visual complexity of architecture. involve the entire physical entity of the built form, and it
According to Vaughan, J. [12], Bovill's method is one of can generate a result which is valid from all points of
only a limited number of quantifiable approaches to view, and represents the real image of the built form
provide a measure of the relative complexity of regardless the differences of viewers.
architectural forms. The approach of Bovill, C. [2] was
based on his assumption that architecture is necessarily
produced through the manipulation of rhythmic forms.
He suggested that fractal geometry can be used to
measure order and surprise.
Bovill's box size: Computational fractal dimension
calculation methods use a range of scaling coefficients
to gradually reduce the grid size and generate more
accurate result. Vaughan, J. [13] discussed the
determination of box size by Bovill, C. [2]. She stated
that Bovill considered the optical action of the eye
focusing from different distances, assuming the location
of the viewer between 40-80 feet away. He then
calculated the approximate grid size for the different
scales, based on the distance of observation and the
angle of the field of view, or the size of the focusing Figure 2. Box-counting dimension
point. of the Robie house [2]
___________________________________________________________________________________________________________

F. Zsuzsanna: “Local Database as Crucial Factor of Sustainable Architecture”, pp. 39–49 41

International Journal of Contemporary Architecture ”The New ARCH“ Vol. 5, No. 1 (2018) ISSN 2198-7688
___________________________________________________________________________________________________________

Most of the problematic parameters within the box D = [log (N(s2)) – log (N(s1))]
Dbb = (5)
counting methods are because it is two-dimensional. [log (s2) – log (s1)]
Such as the lines representation and details levels, which
will be eliminated in a three-dimensional approach. There are two options available to work on a 3D model,
Because a real physical volume will be used. In addition, the first option is to consider the building as a solid
depth, recesses, projection and horizontal layout will be volume, and the second option is to consider the
involved because a full three-dimensional model will be building as an external skin only, which means to count
used. It is mathematically possible to calculate the only the boxes intersect with the external skin. The
fractal dimension using a 3D box counting process for a second option is to consider the building as a solid
3D object, although it does not have a practical volume, and in that case, all the boxes within the
application yet. The same concept of 2-dimensional building space have to be counted. It has been found
plains is applicable on three-dimensional objects. A 3D that it is mathematically valid and more practical to
box counting is mathematically possible, using the same consider the building as a solid volume; because the
equation as in the 2D box counting. external skin thickness will not be mathematically
accurate, and its invisible depth is not architecturally
perceivable.
4.1. Technical difficulties within 3D box
The white space should be determined according to
counting approach
Ostwald, M. [9], which according to his assessment a
50% larger than the object. In a 3D model, the white
A key problem for the box-counting method is that many
space should be increased by 50% in x, y and z-axis.
different scale grid comparisons are needed to get a
Using the same white space will generate comparative
useful or accurate answer. Ideally, at least 13 different
benchmark to validate the 3D box counting approach
scale grids are needed. Bovill, C. [2] does just 2 or 3 grid
based on the validity of the 2D box counting approach.
comparisons, and that's why his results are highly
inaccurate. Some researchers argued that over 25 The cube deletion process
different scale grid comparisons are required. This is The deletion process is suggested as a series of steps on
difficult enough for 2D, but for 3D, it needs a lot of an algorithm for the 3D box counting. This process is the
computing power. easiest way to accurately eliminate the boxes that are
The difficulty in such an approach is that generating 3D not interfering with the building volume, considering the
cubes with the same size over the model will cover the building as a solid entity. This process depends on the
entire model and make invisible even for a computer, deletion of every box once on the row, and if the
unless the cubes were transparent and different in its building volume appeared on the deleted box space, the
digital nature to be differentiated by the computerized deletion process will be undone, and if the volume of the
algorithm or by the bare eyes in a half manual building didn't show up on the deleted box space, the
deleted box will be permanently removed.
application. In this research, only a half-manual
application is about to be performed. Which means that In case of manual box counting, the boxes should be
the building has to be modelled as one object. Then non-transparent, and they should have a contrast
boxes intersects with its skin or located inside it (in case colour, because the process will depend on direct vision.
the building was considered as a solid volume) has to be This process would be performed in order, starting with
detected and counted. On a computer screen, this might one corner, until the top layer is finished, then to the
be problematic, unless a working algorithm or a system next layer, and so on. On a manual work, it will be a
of work is generated and followed. challenge to remember which boxes have been
processed and which are not, so an accurate order has
to be used. It is suggested to work on the top first until
4.2. 3D box counting approach method the top of the building is reached, then to work on each
side, until the façade of the building is reached, then
To perform a three-dimensional box counting, there is a another round of all surface boxes has to be made to
need to generate an algorithm. It should be able to make sure that all the outside boxes has been removed.
generate and analyse boxes through a systematic Then another test has to be done to assure that all the
procedure; then to count the boxes that partially or boxes have been processed, which is to turn the model
totally located within the volume of the building. The to wire frame, to look inside and to check from all sides
process should be taken in several iterations to calculate that there are not a single box free from the building and
the box-counting dimension, using the same formulae of not deleted. The final step is to select the remaining
2D box counting. This fact is mathematically based on boxes by name or by type and command the software to
the one-dimensional denominator for all topological count them. The number of the remaining boxes is (N),
dimensions. and the number of the boxes at the bottom layer of the
___________________________________________________________________________________________________________

F. Zsuzsanna: “Local Database as Crucial Factor of Sustainable Architecture”, pp. 39–49 42

International Journal of Contemporary Architecture ”The New ARCH“ Vol. 5, No. 1 (2018) ISSN 2198-7688
___________________________________________________________________________________________________________

longest side (before deletion) is (e) or (1/s). This process 2D box counting has been used, according to the
should be repeated for several scales, then the following mathematical background of the three-dimensional box
equation has to be -applied for each two scales results: counting formula. The building has been considered as a
solid volume, and the box counting was performed with
the cube deletion process.
Db == [log (N(s2)) – log (N(s1))]
D b (6)
[log (s2) – log (s1)] Iteration 1
The first step is to determine the size of the external box
5. Experiment on Robie house for validation that should contain the grid around the building. The
same Bovill's external box was used, which is 60x24x12
This is an attempt to perform 3 dimensional box meters, which matches the grid size in the two
counting on buildings 3D models, using Autodesk 3ds dimensional box counting of the two elevations of the
Max® program as a three dimensional world. A 3D Robie house, performed by Bovill, [2]. The grid has been
model of the Robie house has been chosen to perform a divided to 3D cubes with 6 meters size, compared with
test application for this approach using almost the same the 7.2 meters grid size of Bovill, but the 6 meters has
grid sizes of Bovill but as three-dimensional cubes, to be been chosen to have simple number without fractions,
able to validate the results using the benchmark of the see Figure 3.
previous studies about the Robie house. This In a shaded view is used to start the cube deletion
experiment was performed manually and the results process, as in the following figure, any cube that was
were between 2.54 and 2.65, which seem to be not only deleted and the building components appears in its
mathematically valid, but also conform to the results of position, will be undeleted, until all the empty boxes are
Bovill, [2] and Vaughan, J. [13]. The same formula of the removed (Figure 4a and b).

Figure 3. Robie house Iteration 1 in 3D (wire frame)

Figure 4. Cube deletion process for Iteration 1

a) Top: Deleting the boxes and retrieving any box that reveals the building
b) Bottom: The boxes inside the volume of the building are also counted
___________________________________________________________________________________________________________

F. Zsuzsanna: “Local Database as Crucial Factor of Sustainable Architecture”, pp. 39–49 43

International Journal of Contemporary Architecture ”The New ARCH“ Vol. 5, No. 1 (2018) ISSN 2198-7688
___________________________________________________________________________________________________________

Figure 5. Robie house Iteration 2

a) Deleting the boxes b) finishing the deletion process

Figure 6. Robie house cube deletion process for Iteration 2

Figure 7. Iteration 2 wire frame check

a) Boxes wire frame b) Boxes forming the shape of the building

Figure 8. Forming the shape of the building

___________________________________________________________________________________________________________

F. Zsuzsanna: “Local Database as Crucial Factor of Sustainable Architecture”, pp. 39–49 44

International Journal of Contemporary Architecture ”The New ARCH“ Vol. 5, No. 1 (2018) ISSN 2198-7688
___________________________________________________________________________________________________________

a) Iteration 3 box formation wire frame model b) Iteration 3 box formation shaded model

Figure 9. The box formation is similar to the original form

Table 1: Robie house 3D box counting

Iteration Total boxes Box size Scale factor (S) Log S Count (N) Log N D(self-semi.)
1 80 6 10 1 50 1.96 1.69
2 640 3 20 1.3 257 2.4 1.84
3 17280 1 60 1.77 4567 3.65 2.06

Iteration 2 Db (box size 6-3) = 2.4-1.69 /1.3-1 = 2.36

The second iteration, the box size became 3 meters Db (box size 6-1) = 3.65-1.69 / 1.77-1 = 2.54
(Figure 5).
Db (box size 3-1) = 3.65- 2.4 / 1.77-1.3 = 2.65 (most
The boxes in iteration 2 are also counted using the cube reasonable)
deletion process (Figure 6).
The next step is to check that all the empty boxes have 6. Validation concept
been deleted using a wire frame model (Figure 7).
In this stage, it will be clear that the boxes starts to form In a box counting method, the results has to be validated
the shape of the building – Figure 8, which relates the by confirming it in two levels of validity, starting by the
visual assessment process to the visual design process. mathematical validity, which requires that the box
counting dimension has to be related to the topological
Iteration 3 dimension; if the box counting was performed two
In the third iteration, the box size is 1 meter, see figure dimensionally, the result has to be between 1 and 2, and
9. Moreover, by applying the cube deletion process, the if it was performed three dimensionally, the results has
resulting form is much more similar to the image of the to be between 2 and 3. If the results were found within
building, and if the iteration continued to the minimum those ranges, it proves that the process is
box size possible, the boxes formation will look exactly mathematically valid, and the equation is applicable.
like a pixilated version of the original form, or even the The second level of validity is to confirm that the
result will be the form itself. The problem in a manual resultant dimension represents the true value of visual
application is that 1 meter or 0.5 meter is the minimum complexity, and this can only be validated by comparing
box size to be workable; the cube deletion process will different samples. In a three dimensional box counting
become impossible in a manual application if the box operation, the results of samples has to be validated in
size is less than 0.5 meters for small buildings, or 1 meter accordance with the results of a two dimensional box
for medium size ones like the Robie house. counting approach for the same samples, the levels of
The remaining boxes are then counted; the computer complexity should be in the same order in both two
will give the result (N), which should be used in the box dimensional and three dimensional box counting
counting equation. Table 1 illustrates the results for operations; this will lead to an architectural problem,
each iteration. that some buildings are complex in some elevations and
very simple in the other elevations, which means that a
Applying the equation: building might be more complex than its opponents in
Db== log (N(s2)) – log (N(s1)) the three dimensional box counting, and yet it has the
D b (7) least visual complexity in two dimensional box counting,
log (s2) – log (s1)
___________________________________________________________________________________________________________

F. Zsuzsanna: “Local Database as Crucial Factor of Sustainable Architecture”, pp. 39–49 45

International Journal of Contemporary Architecture ”The New ARCH“ Vol. 5, No. 1 (2018) ISSN 2198-7688
___________________________________________________________________________________________________________

Figure 10. Single façade complexity

because the chosen façade is the simplest side of the around the sample, and the level of details
building; this problem might affect the validation taken in the sample.
process; therefore, to test the validity of the approach,
4. The mathematical problem represented in the
the chosen samples has to be a single façade buildings,
amount of details within filled boxes as a
the kind of building that have only one complex
statistical problem. And this problem has been
elevation, and the rest of its sides are flat walls, (see the
mathematically neglected by the
building in Figure 10), because in this case, the 2D
mathematicians who established the box
complexity of the front façade will comparatively
counting equations.
represent the same level the 3D complexity of the
building.
For more validity check, a visual preference test to 7.1. The developed approach
determine the experienced visual complexity order by a
normal person might help confirming the box counting 3D box counting is considered a new approach of
results, if the order of complexity calculated by box computational assessment, based on fractal geometry.
counting was the same order which has been chosen by Which delivered mathematically and conceptually valid
normal random selected viewers, considering statistical results in the application on Robie house, using two
validity of the visual preference test. If the approach of major approach parameters: considering the building as
three dimensional box counting method was validated solid volume, and using box elimination approach
through such steps, then it can be computerized in manually (cube deletion process). Figure 11 illustrates
further research, and then it can be used for all types of the parameters of the 3D box counting approach.
buildings to perform further test for its comparability The next step should be to apply this approach on a real
and workability. case, handling an existing built form on multiple scales.
Then to validate the results mathematically and
7. Conclusion geometrically through statistical analysis, then
conceptually and architecturally through visual
preference. This will also help determining the correct
Discussing the fractal assessment tool (box counting) it
has been found that most of its problems are caused by interpretation of results, and to assure that fractal
its two dimensional approach. The problems about 2D dimension represents visual complexity.
box counting were found to be:
1. The conceptual problem represented in 8. Further research
neglecting the layout of the building and the
reassesses and projections. A further development can be performed to this
approach, by involving line segmentation approach to
2. The perceptual problem represented in the scale develop box counting method, using a wire frame
range concept, and the box size perceptual model, based on the concepts of design by lines (Ando,
meaning, and the reflection of human N. 2007). Such development will increase the accuracy
perception on the iterations of the approach.
of 3D box counting on the level of lines representation,
3. The practical problem represented in the line and will reconnect visual assessment process with visual
width of samples, the amount of white space design process. Line count and statistical assessment
___________________________________________________________________________________________________________

F. Zsuzsanna: “Local Database as Crucial Factor of Sustainable Architecture”, pp. 39–49 46

International Journal of Contemporary Architecture ”The New ARCH“ Vol. 5, No. 1 (2018) ISSN 2198-7688
___________________________________________________________________________________________________________

Figure 11. The developed approach

can be involved in such an approach. Also the use of assessment is very important in order to involve its
visual data collection technologies such as laser application in all types of architectural activities, and
scanning and edge detection algorithms will increase the building related tasks, including identity preserving and
workability of the approach. academic research. Such an approach can be developed
to be used even to evaluate students’ projects in the
academic field, and to judge design competitions, easily
8.1. Computerisation of the approach and accurately.
The most important step in further development is
computer application; by programming an algorithm 8.3. Developing computer software for visual
that can perform 3D box counting as developed in this assessment
research and 3D line segmentation. This computerized
algorithm can evolve this computational approach to a Computer software has to be developed to apply the
practical tool. generated visual assessment algorithm. This software
has to perform visual assessment process as the
following:
8.2. Making the approach usable and available
• Provide a 3D environment with the capability to
Computerization of this algorithmic approach with the convert different 3D file formats into suitable 3D
use of data collection technologies, and involving code, and it should be compatible with (3Ds,
computer vision can evolve this approach to become DWG, DXF, ...etc) and all 3D formats that
usable in practice outside the laboratories. Such generated by CAD, BIM or 3D laser extraction.
technology can even be commercially developed and
• When the 3D file is successfully loaded into the
produced in mass scale with the improvement of easy
software interface, a scaling tool must be
user interface, to make available and usable for
available along with measuring tool, in order to
everyone, such as architectural designers, academic
perform scaling correction, and it has to similar
institutions, municipalities and planning authorities, in
addition to heritage preservers. The simplicity of visual
___________________________________________________________________________________________________________

F. Zsuzsanna: “Local Database as Crucial Factor of Sustainable Architecture”, pp. 39–49 47

International Journal of Contemporary Architecture ”The New ARCH“ Vol. 5, No. 1 (2018) ISSN 2198-7688
___________________________________________________________________________________________________________

to AutoCAD tool (scale by reference) to achieve • The software must be able to perform results
maximum scaling accuracy. check and choose a final result by finding the
most frequent result. A statistical analysis tool
• The software must then be able to generate 3D
must be available to re-evaluate the results.
boxes rendered as transparent in a controllable
initial size, multiple iterated sizes with the ability • The software must be able to store the visual
to control the number of iterations, accurate assessment procedure via special file format
overall size (50% larger than model size), and containing all the tables, graphs and 3D data. It
smallest box size to be editable. This should should also be able to compare different file
come with a feature to control the graph of box results and provide statistical reports and graphs
sizes through the iterations to manipulate the based on multiple files.
distribution of the relative box sizes in all • Such software should have a collaborative
iterations. It should also be able to determine (S) interface with BIM software in order to compare
value which is the number of boxes in the X axis. fractal dimension of new designs with existing
• The detection process of which boxes are built form. It will be even suitable to provide a
occupied by the building should follow the cube version of the software via plug-in for BIM
deletion process in order to detect even the programs to allow checking fractal dimension
boxes contain inside the building enclosure, during design.
because the building model will mostly be
hollow, and the findings of the research requires
References
the building to be treated as solid. This means
that only the cubes that don’t encounter the
[1] ANDO, N. (2007). "Design by lines – Study on an
external skin and they are outside the enclosure,
educational method for architectural form
are to be deleted. To achieve this, the software
generation".
should apply an algorithm that detects the
boxes encountered by the skin and the boxes [2] Bovill, C., Wiedemann, G. (1996). Fractal geometry
contained by them. As an alternative, the in architecture and design. Boston, Birkhauser.
algorithm should delete only visible boxes unless
[3] Foroutan-pour, K., Dutilleul, P., Smith, D. (1999).
they are encountered with the model skin (this
Advances in the implementation of the box-
is the algorithm used in the manual application).
counting method of fractal dimension estimation.
All this process must be performed in visible
Applied Mathematics and Computation.
pattern in scene without rendering to minimize
display processor load. [4] Li, J., Du, q., Sun, C. (2009). An improved box-
counting method for image fractal dimension
• The software must then repeat the process in all
istimation. Pattern reconition, volume 42, Issue 11.
iterations and count and record the number of
boxes in each iteration, providing a table of [5] Liang, J., Hu, Y., Sun, H. (2013). The design
values that can be exported via exile format. The evaluation of the green space layout of urban
software must be able to export 3D files for the squares based on fractal theory. Nexus Network
resulting boxes shape of each iteration. This will Journal – Vol.15, No. 1, 2013.
allow manual follow up, and will help the
[6] Mandelbrot, B. (1977). Fractals: Form, chance and
research process by providing detailed data.
dimension. San Francisco, W.H. Freeman and
• The software must not delete the resulting Company.
boxes while following the iterations, it should
[7] Mandelbrot, B. (1983). The Fractal geometry of
rather perform each iteration in an auto-named
nature. W.H. Freeman and Company, New York.
layer; this should allow the software user to re
display each integration of hide it using layer [8] Ostwald, M., Vaughan, J. (2013). Representing
control tools. architecture for fractal analysis: a framework for
identifying significant lines. Architectural science
• The equation of box counting should be applied
review.
through all iterations and between each two
generating fractal dimension values, providing [9] Ostwald, M., Vaughan, J. (2013b). Limits and
all the results in table and graph format. The errors: Optimizing image pre-processing standards
graph format must contain iteration names as for architectural fractal analysis. ARCHITECTURE
on axis and fractal dimension values on the SCIENCE, No.7, pp.1-19, June 2013.
other.
[10] Tucker, C. (2010). Developing computational image
Db = [log (N(s2)) – log (N(s1))] segmentation techniques for the analysis of the
Db =
[log (s2) – log (s1)] visual properties of dwelling facades within a
___________________________________________________________________________________________________________

F. Zsuzsanna: “Local Database as Crucial Factor of Sustainable Architecture”, pp. 39–49 48

International Journal of Contemporary Architecture ”The New ARCH“ Vol. 5, No. 1 (2018) ISSN 2198-7688
___________________________________________________________________________________________________________

streetscape. The University of Newcastle, [13] Vaughan, J., Ostwald, M. (2010b). Refining a
Australia. computational fractal method of analysis (Testing
[11] Tucker, C., Ostwald, M. (2007). Using algorithms to Bovill’s architectural data). Proceedings of the 15th
analyze the visual properties of a buildings style. International Conference on Computer-Aided
AASA Techniques and Technologies: Transfer and Architectural Design Research in Asia CAADRIA
Transformation Conference, UTS. 2010, 29–38. Hong Kong.
[12] Vaughan, J., Ostwald, M. (2010a). A quantitative [14] Vrobel, S. (2011). Fractal time. World scientific
comparison between Wright’s Prairie style and publishing. London.
triangle-plan usonian houses using fractal analysis.
Design principles and practices: An international [15] http://www.circpau.org/content/view/148/402/la
journal. ng,es/

___________________________________________________________________________________________________________

F. Zsuzsanna: “Local Database as Crucial Factor of Sustainable Architecture”, pp. 39–49 49