Digital Visual Exploration Library

Nicholas Tan Jerome and Andreas Kopmann

Institute for Data Processing and Electronics, Karlsruhe Institute of Technology, Eggenstein-Leopoldshafen, Germany

Keywords:

Digital Library, Visual Exploration, Scientiﬁc Visualisation, Information Visualisation.

Abstract:

With the advancement of instrument precision, research facilities are generating data at an unprecedented

rate. These experimental results are stored in a digital library platform which the contents are later acces-

sible from within the facility or the public. However, the sheer volume of collected data is overwhelming

the capacity of researchers and impedes the process of browsing for the desired data. In this paper, we pre-

sent a concept of Digital Visual Exploration Library (DVEL) based on the conﬂuence of two major research

domains—digital library and visualisation—that enables efﬁcient browsing of the growing data within a di-

gital library. We complement the current state-of-the-art textual metadata description by integrating visual

exploration to address big complex data, i.e., data of large size, multimodal data and multivariate data. We

describe our concept based on use cases from three unique domains: climate research with Doppler wind lidar,

X-ray-imaging for entomology research, and medical imaging with ultrasound computer tomography.

1 INTRODUCTION

We are entering an era of data-intensive science where

signiﬁcant data are produced by scientiﬁc experi-

ments worldwide. These data hold the key to verify

scientiﬁc hypothesis and having more data can help

drawing more reliable research conclusions. Hence,

numerous works focus on advancing the instrument

precision which allows researchers to produce data at

an unprecedented rate. For instance, the Large Ha-

dron Collider (LHC) particle accelerator can generate

60 terabytes of data per day (Brumﬁel, 2011). With

LHC serving as an extreme case in the data-intensive

science, such large data are no stranger in most expe-

riments. Often, the data are stored in a digital library

which the contents are accessible from within the re-

search facility or from the public. By storing all the

information into the digital library, the sheer volume

of data complicates the data exploration process. The

main problem lies with users having little knowledge

of the large amounts of data.

To date, many scientiﬁc experiments are starting

to provide open data access to the public (Molloy,

2011) which at the same time enables citizen science

movement—citizens (non-scientists) taking part in

real-world experiments. There is a growing demand

to access scientiﬁc data from the public and impro-

ving the digital library framework is essential. To fa-

cilitate the exploration and analysis of a digital library

collection (Mathew et al., 2017), researchers emp-

hasise metadata analysis, i.e., word-frequency analy-

sis (Shubankar et al., 2011), co-occurrence word ana-

lysis (Isenberg et al., 2017), and probabilistic methods

like Latent Dirichlet Allocation (Grifﬁths and Stey-

vers, 2004). While these methods offer a generic so-

lution, but they fail to provide information related to a

particular domain (Mesbah et al., 2017). Moreover, if

the metadata is not complete or not informative, users

have no other options but to download a copy for off-

line veriﬁcation which is cumbersome.

We propose instead to integrate visualisation

techniques into the framework of a digital library—

Digitial Visual Exploration Library (DVEL). In parti-

cular, we want to utilise visualisation to improve the

data browsing experience. With little knowledge of

the data, the visual presentation in the DVEL enables

users to explore the data faster and with higher con-

ﬁdence. Since the user is the main actor throughout

the exploration process, the interactive interface of the

DVEL allows users to formulate hypothesis while ad-

justing their exploration goals (Keim, 2001). Keim

outlined the beneﬁts of visualisation techniques for

browsing large data sets but has hitherto received little

attention in the digital library domain.

In this paper, we discuss design considerations to

provide appropriate visualisation interface as part of

the DVEL framework. We present three use cases that

illustrate the large data size, multimodal data and mul-

tivariate data. As digital library often provide its data

through a web portal, we are inherently limited by the

Jerome, N. and Kopmann, A.

Digital Visual Exploration Library.

DOI: 10.5220/0006650003410348

In Proceedings of the 13th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2018) - Volume 3: IVAPP, pages

341-348

ISBN: 978-989-758-289-9

341

Table 1: The main components of the web portal for a list of digital libraries in the Physics domain (Last visited 21-07-2017).

Experiment Navigation Preview+Detail Retrieve

NASA Exoplanet Archive (Akeson et al., 2013) List Time Series tbl

Sloan Digital Sky Survey (Eisenstein et al., 2011) Guided search 2D Image jpg

Crystallography Open Database (Gra

zulis et al., 2009) List Jmol (Hanson, 2010) cif

CERN Open Data Portal (South et al., 2011) List — aod

Archie_Nymph

Belbe_01

Belbe_02

Eucrib_01

Eucrib_02

Eucrib_03

Eucrib_04

ASTOR_04_2014

Title Label

LIBRARY

12:45 PM

Aranea

Archie_01

Archie_02

Archie_03

Archie_04

Download

Archie_01

Description

SELECT

PREVIEW

DETAIL

RETRIEVE

NAVIGATION

Archegozetes longisetosus is an oribatid mite (also

known as moss mites), reproduces asexually and has

defensive chemical secretions stored in huge opistho-

notal glands. Unlike other arachnids, oribatid mites

feed on particulate food like dead plant material (e.g.

decaying leaves), fungi and bacteria.

Figure 1: The main components in a web portal of the di-

gital library. Upon selecting a data set, a 3D preview of the

sample is shown together with its details.

bandwidth limitation. Hence, we perform data pro-

cessing based on the data state reference model (Chi,

2000). In particular, we prepare a compact represen-

tation of the data which is small in size and yet best

describe the data.

2 CONCEPT

The ﬁrst use of the term digital library dated back in

1988 by (Kahn and Cerf, 1988), where they presented

an open architecture of a digital library system that

allows convenient access, either locally or remotely,

to geographically distributed users. Back then, most

information is not in digital form, and they described

their architecture as an extension to the traditional li-

brary which users must travel physically to the library

to utilise the resources (Kaur, 2015). However, such

community facility is losing interest from the public

in favour of digital technology advancement—a new

paradigm dubbed the digital library. With informa-

tion resources being digitalised, users are no longer

required to travel physically to a library; but instead,

they can access the information through any digitally

connected medium, i.e. web browser.

Numerous deﬁnitions attempt to deﬁne the digital

library (Cleveland, 1998; Lynch and Garcia-Molina,

1996; Seadle and Greifeneder, 2007; Trivedi, 2010),

which can be broadly distinguished either by its phy-

sical space and collections, by its digital environment,

or by its services. In this paper, we deﬁne library ac-

cording to its primary purpose—a facility which pro-

vides public access to its materials. The term digi-

tal refers to the digital technology which includes the

mobile phone, laptop, computer, network connection,

and cloud service. Hence, a digital library consists

of an extensive collection of digital information with

public access. Similarly, research facilities store their

experiment data in digital form which are accessible

from within the institution or the public.

With open data providing a web portal for remote

data access, users can search and download their de-

sired data conveniently. We identiﬁed the main com-

ponents in the web portal usage based on a set of col-

lective actions to ﬁnd the desired data set: to search

(Navigation), to verify (Preview+Detail), and to do-

wnload (Retrieve). Figure 1 shows the main compo-

nents which are listed as below:

Navigation. A list and a search function to allow data

browsing.

Preview+Detail. A page view presenting details—

usually in tabular format—describing the selected

data set. Optionally, an interactive data preview is

present to emphasise essential features of the data set,

i.e., an interactive 3D visualisation showing the spa-

tial information.

Retrieve. The information of the data set can be

acquired or downloaded in various formats.

We studied four experiments in the physics dom-

ain and discovered that the corresponding web por-

tals mostly provide the Navigation and Retrieve com-

ponents (Table 1). However, the Preview+Detail is

not a prerequisite component and still mainly limi-

ted to textual descriptions. Instead, we highlight the

unique features of each data set by integrating inte-

ractive visualisation techniques such as 3D visualisa-

tion (Levoy, 1988) or information visualisation (Buja

et al., 1991). Although most information can be des-

cribed in texts but conveying complex information so-

lely with texts can be a daunting task. For instance,

an interactive 3D visualisation would provide better

spatial information. Taking the Crystallography Open

Database as an example, the molecular structure is

best described by an interactive 3D visualisation—

Jmol (Hanson, 2010).

To support better data exploration, we propose to

integrate the visual exploration technique as part of

IVAPP 2018 - International Conference on Information Visualization Theory and Applications

342

Table 2: The data stages of each use cases following the data state reference model (Chi, 2000). The Value stage represents

the raw data. The Analytic shows the post-processed data state. The Visual stage depicts the input formats required for the

ﬁnal visualisation (View).

Use case Value Analytic Visual View

NOVA (Sec. 4.1) Image Aggregated Slicemaps Interactive 3D

Database Database

KITcube (Sec. 4.2) Relational Event Summary Encoded 2D Heatmap,

Database Database Images Plan Position Indicator (PPI),

Range Height Indicator (RHI)

USCT (Sec. 4.3) Image Aggregated Encoded 3D Fusion

Database Database Slicemaps

Value

Analytical

Abstraction

Visualisation

Abstraction

View

Data

Transformation

Visualisation

Transformation

Visual Mapping

Transformation

Figure 2: The data state reference model by (Chi, 2000). The black box denotes the state of the data, whereas the arrow shows

the action performed from one state to another.

the Preview+Detail component within the DVEL fra-

mework. The DVEL can provide more information to

complement any shortcomings from the textual meta-

data descriptions. Tominski described the visual ex-

ploration as an undirected search for relevant informa-

tion within the data (Tominski, 2006). Tominski emp-

hasised visual exploration as the process to provide an

overview of the data, to browse through subsets of the

data, and to interact with the data intensively. These

requirements deﬁne what functions a useful visual ex-

ploration tool should have.

Within the DVEL framework, the visualisation

component (Preview+Detail) acts as an add-on fe-

ature to enhance existing solutions. In particular,

the viewing component strives to better inform users

about the data without suffering from the scale of the

massive data.

3 DESIGN STRATEGY

Our approach to design the DVEL revolves around

two main considerations: technology limitations (data

preprocessing) and the effectiveness of data explora-

tion visualisation (three-step process). The techno-

logy limitations refer to the limited network band-

width and the diverse client requirements.

The ﬁnal visualisation should be served at an in-

teractive rate. Hence, data transferred has to be small

and yet still able to represent the original data. Mo-

reover, the transferred data should match the diverse

client requirements, i.e., mobile phone, laptop or de-

sktop. To this end, we perform a data processing stage

which follows the data state reference model (Chi,

2000) (Figure 2). We decided on the type of ﬁnal

visualisation and prepared the required input data ac-

cordingly. We reduce the data for a better network

bandwidth (data reduction) and prepare a hierarchy

of data with different data granularity to serve various

client requirements. Table 2 shows the data stages of

each use cases. The Value stage shows the raw data

input which can be an image database or a relational

database. Then, we perform data processing methods

on the raw data such as data reduction and data fusion.

Depending on the type of ﬁnal visualisation, we trans-

form the data into the required visual formats such as

slicemaps (Tan Jerome et al., 2017b) or encoded ima-

ges (Tan Jerome et al., 2017a).

To design the visual interface in the Pre-

view+Detail component, we involve the domain ex-

perts in formulating primary features of the data

through the human capabilities (Binnig et al., 2016).

This approach allows the visual element to target the

primary users’ needs better. We follow the “informa-

tion seeking mantra” (Shneiderman, 1996) which ser-

ves as our design guidelines—three-step process: (1)

overview ﬁrst, (2) zoom and ﬁlter, and (3) details on

demand. In the overview step, we identify interesting

patterns which best describe the data and enable users

to drill down the details of the data.

4 APPLICATION SCENARIOS

In this section, we discuss the design strategies—data

preprocessing and the three-step process—and appro-

priate visualisation techniques that ﬁt into the Pre-

view+Detail component in three different scenarios:

Digital Visual Exploration Library

343

(1) visualising data of large size (gigabytes of data),

(2) merging multimodal data, and (3) summarising

multivariate data into a single representation.

4.1 Visualising Large Data

The NOVA

project (Schmelzle et al., 2017) serves

as an ideal use case where hundreds of small ani-

mals such as insects and other arthropods are scanned

using the synchrotron-based X-ray microtomography

(SRµCT) imaging technique. These tiny animals are

digitised and turned into large data sets with each size

ranging from 30 GB to 120 GB. The data are stored in

multiple collections of stack images where each stack

of images represents volumetric data. The primary

challenge lies in the size of the data, where transfer-

ring gigabytes of data through the network introduces

high data latency.

Data Preprocessing. To overcome the network li-

mitation, we reduce the data size but at the same

time preserving the quality of the data. As we are

limited by the network bandwidth, we adopted the

multi-resolution slicemap approach to serving vari-

ous clients (Tan Jerome et al., 2017b). The slicemap

is a 3D data structure in the form of a mosaic-format

image, which comprises a series of cross-section ima-

ges. For better user experience, we ﬁrst load a low-

resolution slicemap (128

voxels) for instant visua-

lisation and followed by a high-resolution slicemap

(256

/ 512

voxels) loading in the background (Fi-

gure 4: Tier I). However still, the multi-resolution sli-

cemaps are reduced data which the quality is not com-

parable to the image rendered directly from the origi-

nal image. To address this drawback, we utilise the

resource of the High-Performance Computing (HPC)

server to stream images which are rendered directly

from the original data (Figure 4: Tier II).

Three-step Process. In this application, biologists

would like to identify the type of small animals wit-

hout suffering from the high data latency. We present

two approaches: thumbnail preview (additive blen-

ding) and the two-tier approach—surface rendering

on demand. The ﬁrst approach is integrated into the

Navigation component, where we provide a list of

thumbnail previews (Figure 3A). The thumbnail pre-

view is a 3D image that enables biologists to accu-

rately identify the specimen. Since the image pla-

ceholder in the data list is small, we use primitive

techniques for fast render time by downscaling the

data and performing additive blending. To inspect the

data further, we provide an interactive 3D visualisa-

NOVA stands for Network for Online Visualization and

synergistic Analysis of tomographic data

tion with quality comparable to the desktop applica-

tion (Figure 3B).

4.2 Visualising Multivariate Data

In a large-scale experiment, not only the increasing

data size is a concern, but also the complexity of the

data. These data usually are multivariate data con-

sisting of different attributes such as temperature and

humidity (Kehrer and Hauser, 2013). Despite the

data being highly heterogeneous, researchers must,

however, combine these data with various attributes

and formats to formulate and verify their hypothe-

ses. Using the Doppler wind lidar data from the

KITcube mobile observation platform (Kalthoff et al.,

2013), we demonstrate the DVEL which allows mete-

orologists to browse through multivariate daily data

using the Doppler wind lidar interactive visualisa-

tion (Tan Jerome et al., 2017a).

Data Preprocessing. We separate the multivariate

data based on three different measurement types into

the Event Summary Database. Each type consists

of further three attributes, in which we then norma-

lise them with a constant interval, i.e., one second.

We then reduce the size and merge the attributes into

an encoded image (Figure 5). Although the encoded

images are in the 2D format, the type of ﬁnal visuali-

sation determines how the 2D image is mapped, i.e.,

PPI plot maps the image into a circular scheme, and

RHI plot maps the image into a hemisphere plot. The

mapping is performed interactively using the GPU re-

source through the WebGL shader.

Three-step Process. The main objective is to provide

meteorologists with an overview of the daily trend,

in which we integrate three plots to study the wind

velocity: Plan Position Indicator (PPI) scans, Range

Height Indicator (RHI) scans, and Vertical Stare. In

contrast to the previous list view, we design an ag-

gregated view where we project daily information on

the screen display (Figure 6). In the visual analytics,

we use the daily vertical stare plot as the overview

to detect interesting patterns. We also provide hourly

view which shows a more detailed vertical stare, RHI

and PPI plots. By choosing a particular date (Naviga-

tion component), meteorologists can analyse the daily

overview interactively utilising the brushing techni-

que (Spence, 2001) on multiple attributes.

4.3 Visualising Multimodal Data

When multiple sources are present, we describe the

multivariate data as multimodal data. In the me-

dical domain, multimodal data stem from different

imaging techniques such as CT, MRI, or ultrasound

IVAPP 2018 - International Conference on Information Visualization Theory and Applications

344

Figure 3: Three examples of the arthropods in the NOVA project rendered using (A) the additive blending and (B) the surface

rendering approaches: the oribatid mite Archegozetes longisetosus (left column), the box mite Euphthiracarus reticulatus

(middle column) and an unidentiﬁed Diptera species ﬂy (right column).

Server

Data Manager

SLICEMAP

RAW

Cache Selector

Image Rendering

GPU GPU

GPU

3D Data/

2D Slices Stack

Client Hardware

Requirement

Rendered

Image

Client Viewport

Configuration

Aggregated

Raw / Slicemap

TIER I

TIER II

Navigation Preview+Detail Retrieve

Additive Blending

images

Figure 4: The data preparation of the original data for ﬁnal

visualisation at the client side. A set of preview images

rendered using the additive blending approach is provided

to the Navigation component. For the Preview+Detail, a set

of high quality rendered images are streamed to the client.

data (Kehrer and Hauser, 2013).

Server

Multivariate

Data

Data Types /

Viewport Size

Navigation Preview+Detail Retrieve

Vertical Stare

PPI

RHI

Encoded Images

Figure 5: The data preparation of the multivariate data

for ﬁnal visualisation at the client side (Tan Jerome et al.,

2017a). The multivariate data are processed at the server

before serving to the clients.

Using the USCT

project as our use case (Gemmeke

et al., 2017), we have a digital library ﬁlled with mul-

timodal ultrasound data. The USCT project aims to

develop new image methodology for early breast can-

cer detection. The USCT detector produces three dif-

ferent types of images, which are the reﬂection image,

the sound speed image and the attenuation image. The

sound speed and attenuation images allow doctors or

scientists to classify lesions precisely, whereas the re-

ﬂection image contains the information of the breast

USCT stands for 3D Ultrasound Computer Tomo-

graphy.

Digital Visual Exploration Library

345

JULY

28-07-2016

AUGUST

29-07-2016

30-07-2016

31-07-2016

01-08-2016

02-08-2016

03-08-2016

04-08-2016

NAVIGATION

Download

28-07-2016

SELECT

PREVIEW

+ DETAIL

RETRIEVE

DOWNLOAD

0 120

Range Gates

Time (hour)

Distance (Km)

PPI RHI

Daily Vertical Stare

Hourly Switch

Figure 6: An aggregated view of the Doppler wind lidar

data where each item on the Navigation component shows

a daily overview of the wind velocity which consists of ver-

tical stare, RHI and PPI plots.

Server

Data Manager

SLICEMAP

RAW

Cache Selector

Multimodal

Data

Client Hardware

Requirement

Aggregated

Raw / Slicemap

Navigation Preview+Detail Retrieve

Data Fusion

Sound Speed

Attenuation

Reflection

Figure 7: The main components in a web portal of the di-

gital library. Upon selecting a data set, a 3D preview of the

sample is shown together with its details.

structure (Hopp et al., 2015).

Data Preprocessing. Similar to the data processing

in Section 4.1, we perform data reduction and data

transformation on the original data to produce a se-

ries of slicemaps. Also, we encode the three imaging

modalities into the slicemap which results in an enco-

ded slicemaps (data fusion).

Three-step Process. The goal of this application is

to enable doctors to identify possible tumour regions.

Rather than studying each modality separately, we

fuse the multimodal data and render the ﬁnal image

using the normal fusion approach (Stokking et al.,

2001). The multimodal data are fused according to

the Ranger’s fusion scheme (Ranger et al., 2012). In

Figure 8 shows the multimodal view of a data set,

where:

• Cancer (orange): Sound image thres-

holded at (1.52 ± 0.03) ×10

−3

kms

−1

and attenuation image thresholded at

(0.16 ± 0.04) × 10

−3

dBcm

−1

• Fibroadenoma (green): Sound image threshol-

ded at (1.52 ± 0.03) × 10

−3

kms

−1

and attenu-

ation image not in the thresholded region of

(0.16 ± 0.04) × 10

−3

dBcm

−1

• Fibrous bands or architecture (grey): The whole

reﬂection image range.

DATA

P10

P11

P12

P13

NAVIGATION

Download

P10

SELECT

PREVIEW

+ DETAIL

RETRIEVE

DOWNLOAD

P14

P15

P16

P17

P18

P19

Figure 8: The DVEL for the multimodal USCT breast da-

taset. The orange region shows the area with high sound

speed values suggesting possible tumour tissues. The green

region is the auxiliary attribute (attenuation) to support the

deduction of the tumour possibility.

We transform the fused dataset into 3D formats and

perform surface rendering to obtain an illustration de-

picting the possible tumour region.

5 DISCUSSION AND

CONCLUSIONS

The continuously growing data in a digital library will

pose a challenge in data exploration. Thus, supports

to enable users ﬁnding their data from the exhaustive

list are increasingly important. We proposed Data Vi-

sual Exploration Library (DVEL) that integrates vi-

sual exploration techniques as part of the digital li-

brary framework, where the visual display can repre-

sent the prominent feature of the data better (Keim,

2001). We presented three different use cases and

showed the relevance of our concept with the frame-

work of the digital library.

Due to the bandwidth limitation, we applied a per-

sonalised processing which reduces the memory foot-

print and yet retains the identity of the data. Also,

the smaller data enable a more responsive visual inte-

raction on the client. Furthermore, we utilise the GPU

resource by ofﬂoading graphics and data rendering at

the client. Although image rendered from the origi-

nal large data shows much more detail, the purpose

of our concept is to allow fast data exploration and

ultimately enable users to narrow down their search.

From our experience, the DVEL concept introdu-

ces another type of data to be stored—a compact vi-

sual data. Rather than storing only the original data,

we need to deﬁne a compact visual data which can be

either made of a single attribute or a collection of attri-

butes. The compact visual data has to be designed by

domain experts for better visual communication. The

preparation of the visual data requires close collabo-

ration among researchers of different domains which

might be cumbersome.

Although our use cases differ in its visualisation

purpose, we can set up a design strategy base on the

IVAPP 2018 - International Conference on Information Visualization Theory and Applications

346

data state reference model (Chi, 2000) and the three-

step process (Shneiderman, 1996). However, the cur-

rent state of the design strategy relies heavily on em-

pirical approaches which are not scalable. To this end,

designing a software tool that encompasses the design

strategy proposed can be attractive.

As the implementation of the DVEL in the three

use cases—NOVA (Schmelzle et al., 2017), KIT-

cube (Kalthoff et al., 2013), USCT (Gemmeke et al.,

2017)—are still at its early stage, user feedbacks from

the DVEL remains as an interesting component to

shape the future research challenges. In this paper,

we described mainly on the concept and design of the

DVEL but not the relationship between the meta data

and the visual components.

ACKNOWLEDGEMENT

We would like to thank the German Federal Ministry

of Education and Research for funding of ASTOR and

NOVA (grants 05K13VK5 and 05K16VKB, respecti-

vely). We also express our gratitude to Sebastian

Schmelzle, Michael Heethof, Thomas van de Kamp,

Torsten Hopp, and Michael Zapf for their helpful in-

sights.

REFERENCES

Akeson, R., Chen, X., Ciardi, D., Crane, M., Good, J., Har-

but, M., Jackson, E., Kane, S., Laity, A., Leifer, S.,

et al. (2013). The nasa exoplanet archive: data and

tools for exoplanet research. Publications of the As-

tronomical Society of the Paciﬁc, 125(930):989.

Binnig, C., Fekete, A., and Nandi, A., editors (2016).

HILDA ’16: Proceedings of the Workshop on Human-

In-the-Loop Data Analytics, New York, NY, USA.

ACM.

Brumﬁel, G. (2011). High-energy physics: Down the peta-

byte highway. Nature News, 469(7330):282–283.

Buja, A., McDonald, J. A., Michalak, J., and Stuetzle, W.

(1991). Interactive data visualization using focusing

and linking. In Visualization, 1991. Visualization’91,

Proceedings., IEEE Conference on, pages 156–163.

IEEE.

Chi, E. H.-h. (2000). A taxonomy of visualization techni-

ques using the data state reference model. In Informa-

tion Visualization, 2000. InfoVis 2000. IEEE Sympo-

sium on, pages 69–75. IEEE.

Cleveland, G. (1998). Digital libraries: deﬁnitions, issues

and challenges. IFLA, Universal dataﬂow and tele-

communications core programme.

Eisenstein, D. J., Weinberg, D. H., Agol, E., Aihara, H.,

Prieto, C. A., Anderson, S. F., Arns, J. A., Aubourg,

E., Bailey, S., Balbinot, E., et al. (2011). Sdss-iii:

Massive spectroscopic surveys of the distant universe,

the milky way, and extra-solar planetary systems. The

Astronomical Journal, 142(3):72.

Gemmeke, H., Hopp, T., Zapf, M., Kaiser, C., and Ruiter,

N. V. (2017). 3d ultrasound computer tomography:

Hardware setup, reconstruction methods and ﬁrst cli-

nical results. Nuclear Instruments and Methods in

Physics Research Section A: Accelerators, Spectrome-

ters, Detectors and Associated Equipment.

Gra

zulis, S., Chateigner, D., Downs, R. T., Yokochi, A.

F. T., Quir

os, M., Lutterotti, L., Manakova, E., But-

kus, J., Moeck, P., and Le Bail, A. (2009). Crystal-

lography Open Database – an open-access collection

of crystal structures. Journal of Applied Crystallo-

graphy, 42(4):726–729.

Grifﬁths, T. L. and Steyvers, M. (2004). Finding scientiﬁc

topics. Proceedings of the National academy of Scien-

ces, 101(suppl 1):5228–5235.

Hanson, R. M. (2010). Jmol–a paradigm shift in crystallo-

graphic visualization. Journal of Applied Crystallo-

graphy, 43(5):1250–1260.

Hopp, T., Duric, N., and Ruiter, N. V. (2015). Image fusion

of ultrasound computer tomography volumes with x-

ray mammograms using a biomechanical model ba-

sed 2d/3d registration. Computerized Medical Ima-

ging and Graphics, 40:170–181.

Isenberg, P., Isenberg, T., Sedlmair, M., Chen, J., and

oller, T. (2017). Visualization as seen through its

research paper keywords. IEEE transactions on visu-

alization and computer graphics, 23(1):771–780.

Kahn, R. E. and Cerf, V. G. (1988). An open architecture for

a digital library system and a plan for its development.

Corporation for National Research Initiatives.

Kalthoff, N., Adler, B., Wieser, A., Kohler, M., Tr

aumner,

K., Handwerker, J., Corsmeier, U., Khodayar, S.,

Lambert, D., Kopmann, A., et al. (2013). Kitcube–

a mobile observation platform for convection studies

deployed during hymex. Meteorologische Zeitschrift,

22(6):633–647.

Kaur, H. (2015). Role of Digital Libraries in the Present

Era: Challenges and Issues. In Thanuskodi, S., edi-

tor, Handbook of Research on Inventive Digital Tools

for Collection Management and Development in Mo-

dern Libraries, pages 86–102. IGI Global, Hershey,

PA, USA.

Kehrer, J. and Hauser, H. (2013). Visualization and vi-

sual analysis of multifaceted scientiﬁc data: A sur-

vey. IEEE transactions on visualization and computer

graphics, 19(3):495–513.

Keim, D. A. (2001). Visual exploration of large data sets.

Communications of the ACM, 44(8):38–44.

Levoy, M. (1988). Display of surfaces from volume data.

IEEE Computer graphics and Applications, 8(3):29–

37.

Lynch, C. and Garcia-Molina, H. (1996). Interope-

rability, scaling, and the digital libraries research

agenda. Microcomputers for Information Manage-

ment, 13(2):85–132.

Mathew, G., Agrawal, A., and Menzies, T. (2017). Trends in

topics at se conferences (1993–2013). In Proceedings

Digital Visual Exploration Library

347

of the 39th International Conference on Software En-

gineering Companion, pages 397–398. IEEE Press.

Mesbah, S., Fragkeskos, K., Loﬁ, C., Bozzon, A., and Hou-

ben, G.-J. (2017). Facet embeddings for explorative

analytics in digital libraries. In International Con-

ference on Theory and Practice of Digital Libraries,

pages 86–99. Springer.

Molloy, J. C. (2011). The open knowledge foundation:

open data means better science. PLoS biology,

9(12):e1001195.

Ranger, B., Littrup, P. J., Duric, N., Chandiwala-Mody, P.,

Li, C., Schmidt, S., and Lupinacci, J. (2012). Breast

ultrasound tomography versus mri for clinical display

of anatomy and tumor rendering: Preliminary results.

American Journal of Roentgenology, 198(1):233–239.

Schmelzle, S., Heethoff, M., Heuveline, V., L

osel, P., Bec-

ker, J., Beckmann, F., Schluenzen, F., U. Hammel, J.,

Kopmann, A., Mexner, W., Vogelgesang, M., Tan Je-

rome, N., Betz, O., Beutel, R., Wipﬂer, B., Blanke,

A., Harzsch, S., H

ornig, M., Baumbach, T., and van de

Kamp, T. (2017). The nova project: maximizing beam

time efﬁciency through synergistic analyses of srµct

data. In Developments in X-Ray Tomography XI, vo-

lume 10391, pages 10391 – 10391 – 17. International

Society for Optics and Photonics.

Seadle, M. and Greifeneder, E. (2007). Deﬁning a digital

library. Library Hi Tech, 25(2):169–173.

Shneiderman, B. (1996). The eyes have it: A task by data

type taxonomy for information visualizations. In Vi-

sual Languages, 1996. Proceedings., IEEE Sympo-

sium on, pages 336–343. IEEE.

Shubankar, K., Singh, A., and Pudi, V. (2011). A frequent

keyword-set based algorithm for topic modeling and

clustering of research papers. In Data Mining and

Optimization (DMO), 2011 3rd Conference on, pages

96–102. IEEE.

South, D. M., Group, I. D. S., et al. (2011). Data preser-

vation in high energy physics. In Journal of Physics:

Conference Series, volume 331, page 012005. IOP Pu-

blishing.

Spence, R. (2001). Information visualization, volume 1.

Springer.

Stokking, R., Zuiderveld, K. J., and Viergever, M. A.

(2001). Integrated volume visualization of functional

image data and anatomical surfaces using normal fu-

sion. Human Brain Mapping, 12(4):203–218.

Tan Jerome, N., Chilingaryan, S., Kopmann, A., and Wie-

ser, A. (2017a). An interactive web-based doppler

wind lidar visualisation system. In Workshop on Vi-

sualisation in Environmental Sciences (EnvirVis). The

Eurographics Association.

Tan Jerome, N., Chilingaryan, S., Shkarin, A., Kopmann,

A., Zapf, M., Lizin, A., and Bergmann, T. (2017b).

WAVE: A 3d online previewing framework for big

data archives. In Proceedings of the 12th Internati-

onal Joint Conference on Computer Vision, Imaging

and Computer Graphics Theory and Applications -

Volume 3: IVAPP, (VISIGRAPP 2017), pages 152–

163. INSTICC, ScitePress.

Tominski, C. (2006). Event based visualization for user

centered visual analysis. PhD thesis, University of

Rostock.

Trivedi, M. (2010). Digital libraries: functionality, usa-

bility, and accessibility. Library Philosophy and

Practice (e-journal), page 381.

IVAPP 2018 - International Conference on Information Visualization Theory and Applications

348