High Resolution Images

High Resolution Images
High Resolution Images
Anthony Hamber
123
During the 1980s a nurober of innovatory electronic digital image input, storage, display
and output devices came onto the commercial market. While some of these digital
components have made their way into TV quality domestic electronic products, most
notably the charge couple device (CCD) used in video Camcorders, the more powerful
equipment remained expensive and bad a limited nurober of users. Nevertheless, electronic
digital imaging has already shown that it has very significant implications for those
who wish to study visual data within a wide range of historical contexts and a nurober
of pioneering electronic image projects took place during the late 1980s and early 1990s.1
This chapter will examine the context of high resolution digital imaging raise some of the
issues which will be central to future developments.
Such have been the rapid developments in the enabling technologies that even the
Standard personal Computer is already moving away from merely being able to handle
simple graphics (currently the standard is VGA format) to being able to handle „fullcolour“
images at a reasonable spatial resolution (perhaps in the order of 800×600 pixel).
In the PC world the advent of Windows 3.0, with its ability to handle sophisticated truecolour
graphics, has been another part of the perennial catalyst of the pull of technology
and the push of market forces. lt is therefore clear that a new era in imaging technology
is about to begin. However, those who are concerned with the rapidly increasing demands
of hard disc space on their PCs made by new operating systems (OS2 v.2.0), interfaces
(Windows 3.1) or programs, and are yet to discover the joys of digital imaging, are in a for
a substantial surprise. Digital images can take up very !arge amounts of hard disc space
and the provision of storage for !arge numbers of digital images is key issue which must be
addressed at an early stage of any project.
The high resolution image defies precise definition. Those who attempt such definition
find themselves either making comparison with existing imaging systems, such as
photography and television, or use the capabilities, or more pertently the limitations, of
the existing electronic hardware and software base. While, when compared with a 10 x
Bin transparency, one would not call a 35mm slide a „high resolution“ image the digital
equivalent of 35mm slide would, currently, be thus described. These analogies with the
physical attributes and functionalities of an existing imaging system such as photography
introduce the concept of the current vs. the future capabilities of electronic digital imaging
systems. One of the key elements to this concept is the idea of „future-proofness“ whereby
digital image data originated today will not only be compatible with future systems but
will also be sufficient to answer the needs of future user-requirements, requirements which
are being constantly redefined and expanded as both users and technologists come to terms
with the application of information technology.
The very issue of the application of high resolution digital imaging within historical
research is a complex one. Again the use of digital images within arts teaching and research
1 See ITEM (Image Technology in European Museums and art galleries database), lssue 2,
Europeau Visual Arts Centre at lpswich, 1991 for a Iist of the visual arts projects.
124 Anthony Hamber
is restricted to either existing conventions or a tendency to head of into the realms of Star
Wa.rs where functions may be tcchnically possible but economically not viable. However,
it is absolutely essential that the use of proposed high resolution images is detailed before
imaging takes place. It is only when these requirements have been evaluated by importance
and feasibility that a decision can be made about the necessary spatial resolution and colour
bit depth that should be allocated to each high resolution digital image. In some cases
it will be necessary to carry out detailed research to discover optimum resolutions for
specific functions. lt is also a fact that different sized views of the same image may also be
necessary. i.e. perhaps a quick Iook, a small general view, a full-screen general view, and a
„high resolution“ image to allow panning and zooming or cut and pasting of details. Such
different sizes of images have very important implications for all aspects of the design and
implementation of a system.
The problematic definition of high resolution imaging is compounded by the high
variation in the perception of the capabilities and application of digital imaging systems.
This can be said to be true for both the technologist and the non-technologist user. Using
the paradigm of photography as a bench mark for the design of digital imaging systems
is a two-edged sword. In one respect digital imaging systems permit far greater access,
interactivity and potential than conventional photography while in other respects simple
existing functions are not yet feasible in the digital world. For instance, projecting, in
sequence, 80 35mm slides mounted in a slide projector carousel is a Standard teaching
practice. In an art history lecture dual projection is the norm. Thus a one hour lecture
might have 160 35mm slides. Each 35mm colour slide is the equivalent of around 25 MByte
of information {the equivalent of a 3200 x 2600 pixel 24 bit image). A full 80 slot slide
carousel would be the equivalent of 2 GByte of data. A 35mm slide mounted in a carousel
projector can be called and displayed in under one second. To be able to access and display
on screen such an amount of digital information would take at least 10 seconds even if the
sequence of images was cashed and, furthermore, it were possible to display simultaneously
all the data, which it is not.2
Two characteristics are central to the definition of the resolution of the digital image.
These are spatial resolution, more loosely referred to as „detail“ , and measured in pixels,
and colour data (or bit depth}.
Spatial Resolution
The creation of a high resolution image is the result of the interrelationship of the
„resolutions“ and characteristics of the constituent parts of an imaging chain which includes
the original object, the input device (camera), the storage device and the displayfoutput
device. One consistent element in this imaging chain is that the „image“ will be examined
by the human visual system and thus most imaging systems, such as photography or
television, have been developed to create reproductions which Iook, to the human eye, like
the original scene.
This is a complex issue since different imaging systems use different techniques to
stimulate the human visual system. A colour photograph works on the principle of re-
2 The The current limiting factor is the spatial resolution of computer monitors. This is
discussed in more detail below.
High Resolution Images 125
ßected light using the theory of subtractive colour mixing whereby white light is fittered
using Cyan, Yellow and Magenta filters to produce a gamut of colours. Photographie
transparency material works an the principle of transmitted and refiected light, again using
the subtractive colour mixing principle. The CRT monitor, used by both television
and computers, works an a principle of additive colour mixing by emitted light. These
different principles result in different characteristics of different imaging systems which
become integral parts of their design and construction.
The primary spatial resolution is that of the original object, say a painting. Even at
this Ievel important decisions have to be made before a digital image can be created since
some users of a digital imaging system (such as painting conservators or scientists) will
wish to have greater spatial resolution than that of the resolving power of the human eye at
its minimum focusing distance.3 Research undertaken on the VASARI project indicated
that scanning an original painting at a resolution of 20 pixel/mm would give suflicient
information for all the target users which ranged from scientists and conservators to art
historians.4 While this scanning resolution may not satisfy all users its importance is
covering as wide a gamut as is economically and practically viable.
These findings run in parallel with both applied and theoretical research into image
quality metrics being carried out by both art historians and imaging scientists. The Getty
Art History Information Programm has carried out subjective tests to try and define
parameters for the image quality of images display􀂼d on CRT monitors. s The work centred
on evaluation of juxtaposed colour or grey scale images of different spatial resolution on a
CRT. The published results are inconclusive and are restricted to CRT display which, a.s
will be indicated below, is just one element in the image chain which creates high resolution
images. The complexity of this issue of the definition of a single total quality metric has
been raised by Jacobson and Axford who have looked at the possibilities of effectively
synthesizing the fundamental attributes of the physical measures of image quality.6
The relationship between the Ievel of spatial resolution which is to be imaged from
an original and the spatial resolution of the imaging input device ( camera) is critical one,
and one which has to be immediately addressed. A colour photographic emulsion can
resolve up to 80 Jine pairs (lp/mm) (one black line next to one white line) which is the
equivalent of 160 pixel/mm. In the case of specialist black and white negative films the
resolving power is far greater up to 500 lp/mm or 1000 pixel/mm. In order to scan an
3 Kirk Martinez and Anthony Hamber „Towards a colorimetric digital image archive for the
visual arts“ Electronic Imaging Applications in Graphie Arts, SPIE Proceedings Volume 1073,
pp.ll4-121.
4 Anthony Hamber „The VASARI Project“ Computers and the History of Art, Volume 1, Part
2, 1991 pp.3-20.
t Michael Ester „Image Quality and Viewer Perception“ Visual Resources, Vol.VI I, 1991,
pp.327-352.
6 R.E. Jacobon and N.R. Axford „Towards Total Quality as a means of Assessment of Imaging
Synem.s &om the Users Point of View“ Electronic Imaging and the Visual Arts, The National
Gallery, London, July 1992 who cite Colour, Tone (contrast), resolution (detail), Sharpness (edges)
and Noise (graininess, electronic).
126 Anthony Hamber
entire original at a set resolution, the resolving power of the imaging device has either to
be able to produce in one exposure a single full view of the object at the set resolution or
a nurober of sub-images may have to be made and then joined (or mosaiced) tagether to
create a full view image.
Each photographic material or film has a set resolving power irrespective of its format.
However, a large format colour photographic material ( 10 x 8 inch) will resolve more detail
of a full view of an object than a 35mm format image of the same material and thus for high
quality reproduction I arge format photography has become a standard working practice. A
large format colour photographic transparency with a resolving power of 80 lp/mm could
create an image the equivalent of 39200 x 31360 pixel. If a full colour (24 bit) digital image
is scanned then the file would take up 3.6 GByte. The National Gallery in London has 10
x 8in colour transparencies of all its 2000 paintings. This is the digital equivalent of 6400
GByte.
The CCD electronic digital cameras which began to appear during the 1980s have, to
date, employed one of three technologies. The first is the linear array whereby a single line
of pixels is moved across the image plane of the camera and a series of exposures are made
to build up a two dimension image. A maximum image size of araund 6000 x 4000 pixel
(the theoretical equivalent of a 35mm colour photographic film) is currently possible with
this technology. For a colour image three scans are required, one for the red band, one for
the green band and one for the blue band. The accuracy of the repositioning device moving
the linear array, usually a Steppe motor, can be inaccurate enough to give an image in
which the three bands are in perfect register.
The second CCD technology is the area array in which the photosensitive elements are
set in a fixed two-dimensional matrix giving a fixed resolution. Currently, the maximum
scanning resolution using this technology is araund 2000 x 2000 pixel. This is primarily
due to the difficulty and cost of manufacturing perfect single area array chips.
The third technology centres on the principle of moving in two-dimensional steps a
low resolution area array CCD over the image plane to build up a high resolution image.
High geometric accuracy is achieved by exploiting the characteristics of Piezo crystal.
By electrically charging Piezo crystals (between which the CCD sensor is mounted) very
accurate 20 movement can be achieved and resolutions of araund 3000 x 2000 pixel have
been achieved with this technology even when images have been created from several
individual camera scans.
In reality, many, if not most, digital images are created not directly from the original
but from a photographic reproduction of the original. 35mm colour transparencies are a
common source from which to originate digital images as are colour photographic prints.
Manufacturers of photographic materials produce published data on the characteristics of
films which include the resolving power. However, these resolution tests are based on test
objects (black and white line charts) which givc the optimum resolving power of a film and
an „average“ scene is unlikely to be recorded at this Ievel.
Some manufacturers of desktop digital colour photographic transparency scanners
have produced scanners which can create an image of over 6000 x 4000 pixel from a 35mm
slide, or the equivalent of the optimum resolving power of a colour photographic film.
High Resolution Images 127
Experiments have shown that few colour transparencies have that Ievel of resolvable detail
and a figure of 3000 x 2000 is a more realistic rule of thumb.
A significant recent development has beeu the different Ievels of spatial resolution
which are used as standard on the Kodak Photo CD optical storage system. Kodak have
selected live resolutions:
Base/16
Base/4
Base
4 Base
16 Base
192 x 128 pixel
384 x 256 pixel
768 X 512
1536 x 1024 pixel
3072 x 2048 pixel
These resolutions are intended to cover a range of displayfoutput devices ranging
from a „quick-look“ image (Base/16), through television (Base) and high quality output
(16Base). Only time will tell whether these resolutions become de facto standards (as
has the 35mm photographic format) and prove to be „future-proof‘. The key issue is
whether the 3072 x 2048 pixel format will become the definition for a „High-Resolution“
digital image. The „Base“ is, as Kodak have indicated, television and this is the same
decision made by Philips for their CD-I format in which TV resolution is the highest
image resolution available. Whether this „Base“ .will cbange to High Definition Television
(HDTV) or to a uniform TV /Computer high definition standard may only become apparent
at the end of this decade. One point worth bearing in mind is that these resolutions cannot
take into account the aspect ratio of the original.
Display technologies and output devices are th.e most tangible part of the image
chain. The cathode ray tube (CRT) monitor still dominates electronic digital image
display though various other promising display technologies, such as liquid crystal displays
(LCD), are making rapid progress and may, in time, become the primary display
technology. However, these new technologies are unlikely to have a significant impact in
Jarge-screen monitors (with over a 12in diagonal) until the end of the decade.
A CRI‘ monitor acts as the „window“ through which one can examine a digital file.
The resolution of CRI‘ monitors is comparatively low when compared with input technologies.
The most advanced CRT monitors have a resolution of around 2000 x 1600 pixel.
Most workstation monitors are near 1200 x 900 pixel while at the resolution of PC VGA
monitors is weil below this.
An important consideration which effects the design of the user-interface for a high
resolution digital image system is the variety of aspect ratios within the image chain. An
original may have any aspect ratio though a general rule of thumb is that any variant
can be displayed within a square. This has become a popular design standard for CRI‘
display of images since whatever the aspect ratio of the image it will alw·ays appear in
the same area of the screen. Unfortunately a square is not the commonly found aspect
ratio of input or storage devices. The 35mm photographic format has an aspect ratio of
3:2 a compromise which allows both „landscape“ and „portrait“ images to be produced.
However it is not so easy to change the aspect ratio of electronically produced images as it is
with conventional photography, which requires software and a monitor that can be swivelled
between Landscape and Portrait viewing modes. Television and computer monitors have
128 Anthony Hamber
an aspect ratio of 5:4 (which matches the old format for motion pictures) while High
Definition Television (HDTV) has an aspect ratio of 16:9 (which is the equivalent of the
current motion picture format). Clearly the current 5:4 aspect ratio most closely matches
the „ideal“ square aspect ratio for image display while the 16:9 ratio of HDTV is the rough
equivalent of two 8:8 squares. While some designers intend to exploit this „double square“
by having images appearing in tandem or one „square“ being used a.s a dialogueftext box,
it can be clearly understood that new techniques of image display will be required for
HDTV.
The spatial resolution of the monitor can often play a crucial role in the definition of
high resolution images. There is no point in scanning a general view of an object at greater
resolution than the resolution of the monitor on which the image is to be displayed if the
only aim is to show a general view. This raises one of the key issue of high resolution digital
imaging, that of future-proof data which is flexible enough to be adapted for changes in
both technologies and functionalities. Higher resolution monitors and graphics card which
will permit high resolution display will undoubtedly appear. Conversely, pan and zooll)
functionalities may be required even though a full-view of an object ( pixel for pixel) may
not be possible.
Another area which requires considerable further research is that of the human factors
involved with the creation and use of digital imaging systems. These factors cover a
wide spectrum ranging from the intuitive nature of the human-computer interface and the
design and evaluation of the utility and efficiency of the functionalities defined by userrequirements
to the psycho-physical effect of viewing a CRT monitor at a close viewing
distance for an extended period of time.
Colour Data
A key issue in high resolution digital imaging is that of colour, and the perceived
quality of colour. Using the additive colour theory the mixture of 256 Ievels (8 bits) for
reach of the three primaries (red, green and blue) it is possible to produce over 16 million
colours which will effectively cover all the colours in the visible part of the electromagnetic
spectrum. Assigning 8 bits for each colour band results in 24 bit imaging systems being
designated „true colour“ .
While the theory behind „true colour“ digital imaging is comparatively Straightforward
the reality is somewhat more problematic. This is due to the great number of elements
which have to be taken into consideration. One ofthe primary complications centres on the
non-linear behaviour of the many elements which make up the image input-storage-display
chain.
1) The spectral sensitivity of the imaging device (CCD chips are notorious for their poor
blue sensitivity)
2) The electrical „noise“ which reduces the nurober of accurate bits of data acquired by
electronic digital cameras. Thus a 10 or even 12 bit camera may be required to get 8
bits of true data.
3) The gamut of colour which can be displayed on the CRT rnonitor. This is affected by
the types of phosphors used by CRT manufacturers.7
7 The gamut of reproducible co1our varies according to different forms of reproduction and
High Resolution Images 129
4) Gamma, or contrast, of the va.rious elements which make up the image chain. i.e. the
cantrast in a painting; the cantrast of the colour photographic film which is the copy
to be digitized; the gamma of the digital input device (camera); the gamma of the
monitor for display; the gamma controls in the software. In some cases where there
is a high cantrast ratio 8 bit systems may be insufficient and 10 or 12 bits may be
required to adequately record the range of luminances in a scene.
5) Not only can an image appear perceptually different when displayed on two different
CRT monitors but different programs running on the same machine can produce
different results. Thus one may use one program such as Photostyler to input and
􀃹correct“ an image only to find that it appears in a significantly different form when
displayed in another program. It ca.n be fa.r from clear why this is happening but one
common cause is the manner in which the different programs use the colour Look-Up
Table (LUT).
Input
There a.re few comrnercially a.vailable digital scanners and cameras which produce
irnages of rnore than 1000 x 1000 pixel resolution a.nd they are expensive. Furthermore,
these devices are slow and it is important to appreciate that creating high resolution digital
images is likely to be a time consuming activity for the foreseeable future. For instance a
test using an Apple Quadra 900 with 20 MByte.of RAM and running Adobe Photoshop
took one hour to scan a 35mm slide on a Nikon LS-3500 slide scanner to create a 53
MByte 24 bit image file. Even taking lower resolution images (with a file size of under 1
MByte) still requires several minutes to ca.rry out calibration, pre-scan and then scanning
functions. In addition, image processing functions such as cropping or sharpening rnay be
required. This not only adds to the time taken to input an image but also ha.s hardware
implications, particula.rly with respect to additional RAM.
The systems requirements for ca.rrying out digital scanning on a PC or Apple Macintosh
a.re considerably in excess of a „standa.rd“ rnachine. An increase in RAM and provision
of sufficient ha.rd disc space a.re essential. Experience has shown that when working with
images of 2000 x 1600 pixel or greater, a PC or Macintosh platform may not have the
power to ea.sily ca.rry out image processing functions. Much of the problern lies in the
limitations of image processing architectures and the bottlenecks caused by elements of
these a.rchitectures, notably the computer bus, image bus and processor.8
High resolution images of may cause cra.shes in PC Windows applications and problems
have also been encountered with an Apple Quadra 900. This ha.s been in spite of system
RAM being 20 MByte. However, while the PC and Apple Macintosh/Quadra have an
supply of off-the-shelf programs such a.s Photostyler and Photoshop little, if any similar
commercially available softwa.re exists for the UNIX workstation ma.rket. The commercially
their individual characteristics. The gamut of a colour film is different from the gamut of a
CRI‘ monitor which is in turn different from the gamut of four colour printing. However these
gamuts do, to a great extent, overlap. Particularly problematic colours are yellows, browns and
purple/magentas.
.
8 See Adrian Clark and Kirk Martinez „lmage-Processing Architectures“ in: Image Processing
(ed. Don Pearson), McGraw-Hill, London, 1991 pp.141-55.
130 Anthony Humber
available image processing software for UNIX machines is primarily lirnited to libraries of
image processing routines. Software has to be written to make these routines run under
a windowing environment, such as X 1 l . Until PC and MacintoshjQuadra software is
upgraded for use on UNIX machines (as has already taken place with popular PC programs
such as dBase and WordPerfect) there will remain a very big gap between the powerful
UNIX and the non-UNIX worlds.
Storage & 􀆉anagen1ent
The immediate, short-term and permanent storage of high resolution digital irnages
poses significant problems. In order to run image processing programs additional RAM is a
prerequisite in any computer system. The ability of a computer system to place the whole
of a high resolution image in RAM or virtual memory is of particular importance when
image acquisition and subsequent processing are being undertaken. A practical example
might be of benefit.
If a 24 bit image of 1000 x 1000 pixel has been scanned from a 35mm slide but appears
to be „unsharp“ when displayed on the CRI‘ it may be decided to carry out some image
processing to „sharpen“ it. Most image-editing programs allow three Ievels of „sharpening“
which are simply referred to as „Sharpen“, „Sharpen More“ and „Sharpen Heavily“ . In
order to make comparison between these functions it will be necessary to create three
„Duplicate“ images. The screen thus has four identical version of the digital image file.
Then the three sharpening functions can be applied, one to each of the duplicate irnages.
When this has been achieved direct comparison can be made between an unsharpened,
a „Sharpened“, a „Sharpened More“ and a „Sharpened Heavily“ image. This has very
significant implications for system RAM.
Currently one of the main stumbling blocks in electronic digital imaging is the plethora
of image file formats. The problern centres not just on the fact that there are a variety
of formats such as TIFF, TARGA, PICT, BMP etc. but that there are variants within
each format. These differences and the precise description of the image file formats are
not clearly described in the documentation issued with commercially available software or
available through online help. Thus one may find that a digital image file created and
stored by one program cannot be imported into another program on the same machine
even though the file formal may appear to be the same. Work being carried out by the
International Standards Organization (ISA SC24) aims to produce an international Image
Processing and Interchange (IPI) standard with an Application Programmer’s Interface
(API) and Image Interchange Format ( I IF).9 However, this may be some way off and
market forces have alrea.dy resulted in the development of commercially available image
file formals conversion programs (such as CAMEO IMAGE), though these too may not
recognise all variants of all image file formats.
The !arge amounts of storage space required for digital images has resulted in a considerable
amount of work being carried out on coding (or image compression) techniques.
The problern here is that again there is a Iack of true compatibility between industry
standards.
9 Adrian F. Clark „An International Standard for Image Processing and Interchange“ IAPR
News/etter, Yolume 14, Nurober 1 1991.
High Resolution Images 131
During the past few years much has been written about an image compression technique
called JPEG, an acronym for the Joint Photographie Experts Group which has been
working on an international Standard algorithm for digital still image compression.10 A sister
Standard, MPEG (Motion Picture Experts Group), is an algorithm for an international
Standard algorithm compressing moving digital images.
At the time of writing (Summer of 1992) JPEG has yet to be fully ratified as an ISO
Standard. MPEG is even further away from this point. However, JPEG compression has
now become a standard feature in many image processing programs running on a variety of
computer platforms. These are versions of JPEG and there remains the possible problern
of incompatibility between the different versions. Thus system and platform independent
JPEG image files is still some way off.
An important consideration in the coding and compression of digital image files is the
difference between lossless and lossy techniques. To date algorithms for lossless coding, in
which a decompressed (or restored) image file is bit for bit identical to the file prior to
compression, has indicated that only a compression ratio of araund 2:1 can be expected.
This is known as JPEG Baseline mode. Lossy compression refers to greater compression
ratios which result in some information being lost during the coding/decoding process.
Furthermore, the JPEG compression ratios are not mathematical ratios but image quality
factors, though using a JPEG compression factor as a mathematical ratio can give a rough
idea of changes in image file size. Thus a JPE
.
G compression quality factor of 10 will
result in the compressed file being roughly 1/10th the size of the original file. A JPEG
compression quality factor of between 10:1 and 20:1 gives a decoded image which the
human eye would find difficult to perceive as being different from the original.
However, the degree of compression is highly subject dependent and an important
consideration in the creation of any digital image database is whether lossless coding is
essential or not. This again is an important element in the creation of future-proof data.
The situation regarding JPEG is compounded by the existence of other digital image
coding techniques which are frequently proprietary. Programs such as Photostyler allow
image files to be stored in LZW compressed format. Kodak uses an encoding technique
called PhotoYCC on Photo CD. The emergence of these different encoding techniques is
problematic for the general user since it is far from easy to discover how one technique
relates to another and, indeed, how compatible or incompatible they might be.
The storage of digital data also involves key issues such as the stability and permanence
of the storage medium and the practical realities of data backup. In spite of the statements
made by manufacturers, it is far from clear what sort of life expectancy one can expect
from the various forms of digital data storage media. Magnetic disc is perhaps the least
reliable of these but its comparatively low cost and high data transfer rates mean that it
remains the primary initial storage medium for digital images. The claims for the archival
stability and permanence of optical storage mcdia such as WORM, CD-ROM etc. vary
enormously rauging from 10 years to as many as 100 years. Given the rate of development
in digital storage media during the past decade it may be that any concern about archival
permanence of optical media is unwarranted in that new storage technologies (such a
10 The draft standard is ISO/IEC 10918-1.
132 .4.nthony Hamber
holography) could make existing technologies redundant by the end of the decade. This is
of little comfort to those who experience catastrophic failure of an archival storage device.
In reality many digital image databases stored on optical media are backed up on magnetic
disc and magnetic tape.
The practical issues involved in creating, storing and managing a high resolution
digital image database are still very considerable. Firstly, there are no fully integrated
imageftext databases. Working within the PC world one must scan an image and store
it as an 8 character DOS file before linking it to a database which can handle images as
external files. Superbase 4 and Toolbook, both Windows 3.0/3.1 applications can handle
24 bit digital images. Superbase 4 treats the image as an external file though images can
be saved in a Form Design. In Toolbook, which bears direct comparison with Hypercard
in terms its use of object-orientated programming, it is possible to save the image on a
card thus integrating the two data sets into a single object. However, this is not the case
with Hypercard which still handles images as external files.
Future image processing programs will require input, processing and imageftext databasing
functions in one seamless and integrated program. The programs could also include
search/retrieval functions based on textual as weil as visual description of image content.11
Given the size and complexity of digital image databases it is likely that in the not too
distant future they will form part of an open, networked, distributed imaging environment
making full use of an object-orientated user interface. Strack and Neumann have recently
outlined such an open imaging environment in which devices (image sources and destinations),
Operators (derived from a commercial image processing library), and images (of
different data types) are managed and presented uniformly to the user.12 However, these
theoretical system designs are quite some way away from the current practial realities of
creating and managing high resolution image databases.
Output
Over recent years there as been a steady rise in the nurober of high quality output
peripheral devices which can be attached to computers. These vary from 35mm slide
writers with 4000 line resolution to thermal transfer or dye sublimation hard copy printers.
All these peripheral devices are currently extremely expensive, one Kodak printer costing
just under .t 20,000. However, the rise of these printers, and the fact that their overall
quality has risen dramatically points, towards the continued need for high quality hard
copy (which in turn implies „high resolution“ input) and the probability that demand for
such devices will quickly reduce prices.
Some of these printers are extremely sophisticated and may include a dedicated CPU
and software which will interpolate the digital data in a „low“ resolution image file to
11 See C.H.C. Leung and J.N.D. Hibler Areruteeure of a Pictorial Database Management System,
Report for tbe Research and Development Department, The Britisb Library, 1991 and
William Vaugban „Paintings by Number: Art History and tbe Digital Image“ in: Computers
and The History of Art (ed. A. Hamber, J. Miles and W. Vaughan), Mansell, London, 1991
pp.74-97.
12 R. Strack and L. Neumann „Object-orientated image database for an open imaging environment.“
Electronic Imaging and the Visual Arts, The National Gallery, London. July 1992
High Resolution Images 133
produce a print which has the full resolution of the printer. Conversely the printer will
sub-sample data from a “high“ resolution image file to reduce the resolution to match the
full resolution of the printer device. This is a process used by Kodak for hard copy output
from the Photo CD format.
The ability to produce limitless number or identical 35mm slides or hard copy prints
from the same digital image file is an exciting prospect for slide librarians having to face a
member of staff or Student who has lost or damaged an irreplaceable image. On the other
band it send shudders down the spine of lawyers, publishers and Copyright holders.
lssues Concerning Copyright
lt may seem somewhat ironic that high resolution digital image databases, their use
and their functionalities may initially be restricted by legal rather than technical considerations.
Two issues are particularly prominent. Firstly is the Ievel of potential piracy permitted
through digital imaging techniques whereby an exact aud easily repeatable copy can be
made of a digital file. Secondly is the complex issue of electronic copyright and reprod uction
right. Work is underway to address both issues and it is essential that academic users of
digital images and digital image databases made themselves heard during the ensuing
debate. In the UK art historians made a fundamental mistake in not being represented in
the discussions which resulted in the 1988 Copyright Act which makes the making of slides
from published material which is still in copyright either impractical or uneconomic. A
similar situation must not be allowed to arise for the use of the digital image for teaching
and research purposes.
Algorithms to prevent illegal copying are being developed and an EEC project named
CITED – Copyright in Transmitted Electronic Documents · (ESPRIT 11 Project No.5469)
is examining ways of safeguarding copyright material that is stored and transmitted in
digital form.
The issue of electronic reproduction right and reproduction fees is one which is difficult
to clearly define due to the immature nature of the market and the fundamental differences
between electronic and traditional publishing. Even the most simple of questions has yet
to be addressed. For instance, is an academic institution legally entitled to turn its 35mm
slide library into a digital image database for teaching and research? If not, what is
the situation regarding reproduction fees? ls the fee dependant on the spatial resolution
of the digital image? On the colour bit depth? On the number of times the image is
retrieved/displayed? On a flat fee in the form of a one off payment? On an annual fee?
What is the situation regarding including digital images in students essays or programs?
What is the situation regarding making hard copy print outs?
These questions also interrelate with the issue of the electronic publishing of texts,
graphics, and audio to produce multimedia applications, whether they be commercial or
otherwise.
134
Conclusion
Anthony Hamber
High resolution digital imaging is now a practical and economic reality. Yet much work
needs to be done to help nurture this new field. Standards need to be clearly defined to help
the holders of the information, the originators of the digital images and the users create
an environment in which the potential of the digital image can be effectively exploited for
wide variety of commercial and educational applications. The experience of the past ten
years has shown that user involvement is a key element in the evolution of systems and
the commercial markets which will make them economically affordable. The user can not
only play a pivotal rote in the design of systems software and hardware but also in shaping
the forms and functions of the application of high resolution digital image.
Halbgraue Reihe
zur Historischen Fachinformatik
Herausgegeben von
Manfred Thaller
Max-Planck-Institut für Geschichte
Serie A: Historische Quellenkunden
Band 14
Erscheint gleichzeitig als:
MEDIUM AEVUM QUOTIDIANUM
HERAUSGEGEBEN VON GERHARD JARITZ
26
Manfred Thaller (Ed.)
Images and Manuscripts
in Historical Computing
Max-Planck-Institut für Geschichte
In Kommission bei
SCRIPTA MERCATURAE VERLAG
St. Katharinen, 1992
© Max-Planck-Institut für Geschichte, Göttingen 1992
Printed in Cermany
Druck: Konrad Pachnicke, Göttingen
Umschlaggestaltung: Basta Werbeagentur, Göttingen
ISBN: 3-928134-53-1
Table of Contents
lntroduction
Manfred Tb aller . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
I. Basic Definitions
Image Processing and the (Art) Historical Discipline
.Jörgen van den Berg, Hans Brandhorst and Peter van Huisstede ……………. , .. 5
II. Methodological Opinions
The Processing of Manuscripts
Manfred Tballer …….. . ……….. . . . … …. . . .. . . . ……………… . . .. …… 41
Pietonal Information Systems and the Teaching Imperative
Frank Colson and Wendy Hall . . . . . . . . . . . . . . . . . . . . . .. . .. . . .. … . . . . . .. . . . . . . . . . . . 73
The Open System Approach to Pictorial Information Systems
Wendy Hall and Frank Colson . . . . . . . . . . . . . . . . . . . . . . . . . . . . … . …….. . . . . . . .. . . . 87
111. Projects and Case Studies
Tbe Digital Processing of Images in Archives and Libraries
Pedro Gonzi.lez . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
High Resolution Images
Anthony Hamber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . .. . . . . . . . . . . . . . . . 123
A Supra-institutional Infrastructure for Image Processing in the Humanities?
Espen S. Ore . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Describing the Indescribable
Gerhard Jaritz and Barbara Schub . . . . . …. . … . . . . . . . . . . . . . . … . . . . . . .. . . . . .. . 143
Full Text / Image DBMSs
Robert Rowland . . . . . .. . …. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
lntrosluctjon
lntroduction
Manfred Thaller
This book is the product of a workshop held at the International University Institute
in Firenze on November 151h, 1991. The intention of that workshop has been to bring
tagether people from as ma.ny different approaches to „ima.ge processing“ as possible.
The reason for this „collecting“ approach to the subject was a feeling, tha.t wbile image
processing in many ways has been the „hattest“ topic in Huma.nities computing 1n recent
years, it may be the least weil defined. It seems also much barder to say in this area., wbat
is specifically important to historia.ns, tha.n to other people. In that situation it was feit,
that a foruin would be helpful, which could sort out what of the various approaches can
be useful in historical resea.rch.
To solve this task, the present volume has been produced: in ma.ny ways, it reflects
the discussions which actually have been going on less, than the two compa.nion volumes
on the workshops at Glasgow a.nd TromS0 do. This is intentional. On the one band,
the pa.rticipa.nts at the workshop in Firenze did strongly feel the need to have projects
represented in the volume, which were not actually present at the workshop. On the other,
the discussions for quite some time were engaged in cla.rifying what the metbodological
issues were. That is: what a.ctua.lly a.re the topics for schola.rly discussion beyond the
description of individual projects, when it comes to the processing of images in historical
resea.rch?
The situation in the a.rea is made difficult, because some of the underlying a.ssumptions
are connected with vigoraus research groups, who use fora of schola.rly debate, which are
only slightly overlapping; so, what is ta.citly a.ssumed to hold true in one group of research
projects may be considered so obviously wrang in a.nother one, that it sca.rcely deserves
explicit refutation.
We hope, that we have been succes:.ful in bringing some of these hidden diJferences
in opinion out into the open. We consider this extremely importa.nt, because only that
cla.rification allows for a fair evaluation of projects which may have sta.rted from different
sets of a.ssumption. So importa.nt, indeed, that we would like to catalogue here some of the
basic differences of opinion which exist between image processing projects. Tbe reader will
rediscover them in many of the contributions; as editor I think however, that suma.rizing
tbem at tbe beginning may make the contributions- which, of course, have been striving
for impartiality – more easily rccognizable as parts of one coherent debate.
Three basic differences in opinion seem to exist today:
(1) Is ima.ge processing a genuine and independent field of Computer ba.sed resea.rcb in
the Humanities, or is it an auxiliary tool“? Many projects a.ssume tacitly – a.nd some do so
quite outspokenly- that imag􀃣 on the computer act as illustrations to more conventional
applications. To retrieval systems, as illustrations in catalogues and the like. Projects of
this type tend to point out, that with currently easily available equipment a.nd currently
clearly understood data processing technologies, the analysis of images, which can quite
easily be ha.ndled as illustrations today, is still costly and of uncertain promise. Wb ich is the
rea.son why they a.ssume, that such analytical approaches, if at all, should be undertaken
2 Introductjon
as side effects of projects only, which focus upon the relatively simple administration of
images. Their opponents think, in a nutshell, that while experiments may be needed, their
overalJ outcome is so promising, that even the more simple techniques of today should be
implemented only, if they can later be made useful for the advanced techniques now only
partially feasible.
(2) Connected to this is another conflict, which might be the most constant one
in Humanities data processing during the last decades, is particularly decisive, however,
when it comes to image processing . Shall we concentrate on Ievels of sopbistication, which
are available for many on today’s equipment or shall we try to make use of the most
sophisticated tools today, trusting that they will become available to an increa.singly !arge
number of projects in the future? This specific battle has been fought since the earliest
years of Humanities computing, and this editor has found bimself on both sides at different
stages. A „right􀅁 answer does not exist: the debate in image processing is probably one
of the best occassions to understand mutually, that both positions are full of merit. It is
pointless to take permanently restrictions into consideration, which obviously will cease to
exist a few years from now. It discredits all of us, if computing in history always promises
results only on next years equipment and does not deliver here and now. Maybe, that is
indeed one of the more important tasks of the Association for History and Computing:
to provide a link between both worlds, Jending vision to those of us burdened down by
the next funding deadline and disciplining the loftier projects by the question of when
sometbing will be affordable for all of us.
(3) The third major underlying difference is inherently connected to the previous ones.
An image as such is beautiful, but not very useful, before it is connected to a description.
Shall such descriptions be arbitrary, formulated in the traditionally clouded langnage of
a historian, perfectly unsuitable for any sophisticated technique of retrieval, maybe not
even unambigously understandable to a fellow historian? Or shall they follow a predefined
catalogue of narrow criteria, using a carefully controlled vocabulary, for both of which it is
somewbat unclear how they will remain relevant for future research questions which have
not been asked so far? – All the contributors to this volume have been much to polite to
pbrase their opinions in this way: scarcely any of them does not have a strong one with
regard to this problem.
More questions than answers. „Image processing“, whether applied to images proper
or to digitalized manuscripts, seems indeed to be an area, where many methodological
questions remain open. Besides that, interestingly, it seems to be one of the most consequential
ones: a project like the digitalization of the Archivo General de Indias will
continue to influence the conditions of historical work for decades in the next century.
There are not only many open questions, it is worthwhile and neccessary to discuss them.
While everybody seems to have encountered image processing in one form or the
other already, precise knowledge about it seems to be relatively scarce. The volume starts,
therefore, with a general introduction into the field by· J. v.d. Berg, H. Brandhorst and
P. v. Huisstede. While most of the following contributions have been written to be as self
supporting as possible, this introduction attempts to give all readers, particularly those
lntroductjon 3
with only a vague notion of the techniques coucerned, a common ground upon which the
more specialized discussions may build.
The contributions that follow have been written to introduce specific areas, where
handling of images is useful and can be integrated into a !arger context. All authors have
been asked in this part to clearly state their own opinion, to produce clearcut statements
about their methodological position in the discussions described above. Originally, four
contributions were planned: the first one, discussing whether the more advanced techniques
of image processing can change the way in which images are analysed and handled by art historians, could unfortunately not be included in this volume due to printing deadlines:
we hope to present it as part of follow up volumes or in one of the next issues of History
and Computing.
The paper of M. Thaller argues that scanning and presenting corpora of manuscripts
on a work station can (a) save the originals, (b) iutroduce new methods for palaeographic
training into university teaching, (c) provide tools for reading damaged manuscripts, the
comparison of band writing and general palaeographic studies. He further proposes to
build upon that a new understanding of editorial work. A fairly long tr.chnical discussion
of the mechanisms needed to link images and transcriptions of manuscripts in a wider
context follows. ·
F. Colson and W. Hall discuss the role of images in teaching systems in university
education. They do so by a detailed description of the mechanism by which images are
integrated into Microcosm I HiDES teaching packages. Their considerations include the
treatment of moving images; furthermore tbey enquire about relationships between image
and text in typical stages in the dialogue between a teaching package and a user.
W. Hall and F. Colson argue in the final contribution to thill part the general case
of open systems, exemplifying their argument with a discussion of the various degrees in
which control about the choices a user has is ascertained in the ways in which navigation is
supported in a hyper-text oriented system containing images. In a outshell the difference
between „open“ and closed systems can be understood as the following: in an „open
system“ the user can dynamically develop further the behaviour of an image-based or
image-related system. On the contrary in static „editions“ the editor has absolute control,
the user none.
Following these general description of approaches, in the third part, several international
projects are presented, which describe in detail the decisions taken in implementing
„real“ image processing based applications, some of them of almost frigthening magnitude.
The contributors of this part were asked to provide a different kind of introduction to the
subject than those to the previous two: all of them should discuss a relatively small topic,
which, however, should be discussed with much greater detail than the relatively broad
overviews of the first two parts.
All the contributions growing out of the workshop came from projects, which had
among their aims the immediate applicability of the tools developed within the next 12-
24 months. As a result they are focusing on corpora not much beyond 20.000 (color) and
100.000 (blw) images, which are supposed to be stored in resolutions manageable within
:::; 5MB I image (color) and :::; 0.5 MB I image (blw). The participants of the workshop
feit strongly, that this view should be augmented by a description of the rationale behind
4 lntroductjon
the creation of a !arge scale projt’Ct for the systematic conversion of a complete archive.
The resulting paper, by P. Gonza!ez, describes the considerations which Iead to the design
of the .’\rchivo General de Indias projt’Ct and the experiences gained du ring the completed
stages. That description is enhanced by a discussion of the stratrgies selected to make the
raw bitmaps accessible via suitable descriptions I transcriptions I keywords. A critical
appraisal, which decisions would be made dilferently after the developments in hardware
tecbnology in recent years, augments the value of the description.
The participants of the workshop feit furthermore strongly, that their view described
above sbould be augmented by a description of the techniques used for the handling of
images in extremely high resolution. A. Hamber’s contribution, dealing with the Vasari
project, gives a very thorough introduction into the technical problems rncountered in
handling images of extremely high quality and also explains the economic rationale behind
an approach to start on purpose with the highest quality of images available today on
prototypical hardware.
As these huge projects both were related to iustitutions which traditionally collect
source material for historical studies, it seemed wise to include also a view on the roJe
images would play in the data archives which traditionally have been of much importance
in the considerations of the AHC. E.S. Ore discusses what implications this type of machine
readable material should bave for tbe infrastructure of institutions specifically dedicated
to Humanities computing.
Image systems which deal with the archiving of pictorial material and manuscript
systems have so far generally fairly „shallow“ descriptions. At least in art history, moreover,
the rely quite frequently on pre-defint’d terminologies. G. Jaritz and 8. Schuh describe
how far and wby historical research needs a different approacb to grasp as much of the
intemal structure and the content of an image as possible.
Last not least R. Rowland, who acted as host of the workshop at Firenze, describes tbe
considerations which currently prepare the creation of another largescale archival database,
to contain !arge arnounts of material from the archives of the inquisition in Portugal. His
contribution tries to explore the way in which the more recent developments of image
processing can be embedded in the general services required for an archival system.
This series of workshop reports shall attempt to providr a broader basis for thorough
discussions of current methodological questions. ‚fheir main virtue shall be, that
it is produced sufficiently quick to become available, before developments in this field of
extremely quick development make them obsolete. We hope we have reached that goal:
the editor has to apologize, however, that due to the necessity to bring this volume out in
time, proofreading has by neccessity be not as intensive as it should have been. To which
􀂰nother shortcoming is added: none of the persons engaged in the final production of this
volume is a native speaker of English; so while we hope to have kept to the standards of
what might be described as „International“ or „Conti111mtal“ English, the native speakers
among the readers can only be asked for their tol(‚rance.
Göttingrn, August 1992