`
`Technical
`Association
`of the
`Graphic
`Arts
`
`TAGA
`
`"Disseminating
`graphic arts
`research
`internationally
`since 1948."
`
`68 lomb Memorial Dr
`Rochester. NY 14623-5604
`
`
`
`The 1997 T AGA Proceedings was edited and assembled by
`the T AGA Office, Rochester, New York and printed by
`BookCrafters, Inc. of Chelsea, Michigan.
`
`
`
`P ublication Rights Agreement
`
`The Technical Association of the Graphic Arts (TAGA) is pleased to consider the
`publication of your Work. tentatively entitled:
`
`Title: _____________________ _
`
`to be included in the following publication: TAGA Proceedings.
`
`G ra nt of Righ ts: As a condition of publication, you hereby grant to T AGA the
`following rights: 1) the exclusive right of first publication of the Work throughout the
`world as part of the publication named above; 2) the non-exclusive right to reprint the
`Work whenever necessary in any medium or form of communication; and 3) the right to
`use the Work, or any part thereof, in any other publication produced by T AGA. You
`shall reserve all exclusive rights not specifically granted to T AGA herein and will httve
`the right, after the Work hns been published. to reprint the Work in any publication.
`provided that you include in the publict~tion the proper credit to TAGA for prior
`publication of the Work.
`
`W arranty: You wammt that the Work is original with you. that its public:uion will not
`infringe on the rights of others, and thai you hnve fu ll power to make this agreement.
`
`Previous P u blication a nd Permission: You warrant that the Work has not been
`published elsewhere in whole or in part (except as may be set out in a rider below) and
`thnt no agreement to publish the Work or any part or version thereof is outstanding.
`Should the Work contain any material which requires written permission for inclusion
`in the Work. you agree to obtain such permission from the copyright proprietor.
`
`If the foregoing terms are satisfactory. please sign nnd date this agreement below and
`return it to the project coordinator or editor at the above address.
`
`Author(s):
`
`Signature
`
`D<tle
`
`U.S. Government Employees. Please check (a) or (b):
`
`(a)
`
`(b)
`
`This Work was written on my own time and not required by my
`assigned job or official duties as a U.S. Government employee.
`
`This Work was written as part of my assigned job or official duties
`as a U.S. Government employee.
`
`
`
`A Framework for Digital Data Workflow
`in a Graphic Arts System
`
`Robert R. Buckley*
`
`Keywords: prepress, systems. digital. exchange. flow
`
`Abstract: A basic desktop publishing system connects the application or creative
`operation to an output device via a driver and RIP. As the desktop has become a
`more credible source of input for graphic arts and production systems. the
`demands on productivity, quality, expertise and integration with existing processes
`has grown. One result has been a more elaborate framework for digital processing
`operations from creation to output, as described in this paper. Along with the
`proliferation of processing stages has come a growth in commercial opportunities,
`standardization activities and workflow scenarios. The usc and application of this
`framework are illustrated by examples drawn from existing commercial systems.
`
`Introduction
`
`In a graphic arts system, a creative source is linked to a production
`destination. The basic elements of such a system can serve as the starting point for
`describing the evolution of digital data processing and flow in graphic arts
`systems. These basic elements are shown in Figure I, where an application is
`connected to an output device via a print driver and RIP (Raster Image Processor).
`
`The application is feature-rich editing software with a good user interface
`for capturing user intent and generating the desired document appearance. which is
`expressed in terms of an internal format. lb.e driver converts the internal format to
`a PDL or page description language. such as PostScript or PDF (Portable
`Document Format). Most high end graphic arts applications have integrated
`drivers. The RIP converts the device-independent PDL into a device-dependent
`raster image for the target output device. The output device can be an imagesctter.
`direct to plate, direct to press, or a laser printer. While the physical medium can
`influence the overall workflow, this paper will focus on the digital data flow and
`operations in the system.
`
`*Digital Imaging Technology Center. Xerox Corp .. Webster, NY
`
`337
`
`
`
`Designer
`
`Output Medium
`
`Figure I: Basic Framework
`
`Taken together, these steps fom1 the core of a commercial system. Since
`they precede the output, they are "prepress" operations. Adding the necessary
`infrastructure, including platforms, networks, storage and databases would turn
`the framework of Figure 1 into a commercially viable system.
`
`In this basic system, the application is the value added operation. The
`remaining steps are essentially format converters. They translate and interpret the
`content choices made by the document designer at the application into formats
`appropriate to the different stages in the tlow. In the progression from application
`to output device, the ability to modify document content decreases, while data
`volume and dependence on output device characteristics increase. For example, the
`application's internal format describes a document compactly in terms of objects,
`such as lines, figures. characters, images and so on. By comparison. the image
`format describes the document in terms of pixels that are tuned to the output
`device's resolution, inks and operating behavior, possibly including compensation
`for characteristics such as misregistration and tone reproduction.
`
`Multiple Applications and Outputs
`
`While Figure 1 shows a vertical, one dimensional tlow, the system
`designer has a choice of applications and output devices. depending on the nature
`of the job, the destination and throughput requirements. The system designer may
`configure the system to offer multiple applications, with integrated or shared
`drivers, and multiple outputs, each usually with a dedicated RIP (Figure 2).
`
`The enabler for such a horizontal, two dimensional system is the PDJ .,
`usually PostScript, which is the common intermediate format that a variety of
`different applications can use to connect and communicate with a variety of
`
`338
`
`
`
`different devices. Using PDL theoretically decouples applications and output
`devices, so that an individual application or device can be added. updated or
`deleted without affecting the rest of the system.
`
`One of the output devices in Figure 2 would be used for production, while
`another would be used for proofing. In this case, the designer (or operator in U1e
`prcpress environment) runs through the steps to a proofer, whose output provides
`feedback, indicating where changes and corrections arc needed, and something for
`the client to approve, before committing to a production run. Presumably the
`proofer' s output is an inexpensive, accessible simulator of the production output.
`
`Designer/Operator
`
`RIP
`
`Dev
`
`RIP
`
`Oev
`
`Proofer
`
`Output
`
`Figure 2: Multiply connected system
`
`The proofer is part of a feedback loop shown in Pigure 2 that
`encompasses the entire process from application through RIP to output device. It
`is desirable to make the feedback loop tighter and faster, either with soft proollng.
`by having the proofer co-located with the designer or client, or by making the
`proof-correction cycle encompass fewer steps in the process. The latter implies
`that the ability to make corrections to the digital data exists other places than the
`application. lllis is especially important in the production environment. which is
`typically separated from the designer and the application, so that it is awkward
`and time consuming to have to go back to the application every time a change is
`needed.
`
`Related to the idea of feedback and correction is the notion of binding.
`Binding is making and committing to output-specific choices in the digital data
`content. Examples of binding choices are which inks to use, which is determined
`by the target output device, and what text to insert. especially when the text is
`variable data.
`
`Early binding means decisions, selections and corrections are made in the
`application or driver, as part of the design stage. where the most skill is usually
`
`339
`
`
`
`available. An example of early binding is specifying traps in an application. such
`as QuarkXPress. In the basic digital data flow shown in Figure I, there is often no
`other place to make some binding decisions. Early binding can make RIPping a
`simpler process, but it makes the PDL file more output specific and less portable.
`lllis means that redirecting a job to a different output device happens back at the
`application.
`
`Late binding delays output-specific decisions to the last possible moment.
`lllis improves PDL device independence and portability. It also means that
`modifications do not require going back to the application and in effect starting
`over, which would reduce productivity. But late binding places greater demands on
`the capabilities and skills needed later in the digital data tlow. Besides where one
`does the binding, how one does it is also an issue, and specifically whether it is
`interactive or automated.
`
`Prepress Systems
`
`The basic system in Figure 1, typical of desktop applications, has evolved
`into the more elaborate framework shown in Figure 3, which applies to prepress
`systems in graphic arts applications. 'The transition from one to the other has
`added operations before, after and within the RIP. lllis means added opportunities
`in the system for changing content and adding value to the final printed product.
`Several forces have driven this transition:
`
`•
`
`Improved productivity
`- greater throughput
`- faster turnaround
`- more predictability
`- reduced rework
`- more automation
`Increased connectivity
`•
`Improved quality
`•
`• Reduced cost
`• Enhanced capabilities
`
`Improved productivity has several components, from getting a job through
`the system and into the client's or customer's hands more quickly, to being able to
`having the system operate in a more predictable or reliable way in the face of the
`vagaries of input and output. Other ways of improving productivity are reducing
`the rework and backtracking when something does not come our right, or does nor
`come out at all, and looking for places where processes can be automated, thereby
`reducing the need for interactions and decisions. which take time, energy and skill,
`and increase cost.
`
`340
`
`
`
`lhere is the continual demand to reduce cost, improve quality and meet
`customer requirements and expectations. Another force is improving connectivity
`on both the input and output sides. Enhanced capabilities and new features create
`new product opportunities and services, attract business and provide an outlet for
`innovation.
`
`Designer
`I
`Application
`
`Print Driver
`I PDL
`I PreProcessor I
`I PDL
`~RIP Interpreter
`1 Displa y List :
`Filters
`
`Imager
`1 Image Format
`I PostProcessor I
`Image Format
`1
`Modulator
`
`Device
`
`I
`
`Output Medium
`Figure 3: Enhanced Framework for Prepress Operations
`
`Pre-RIP Processing
`
`When files are transferred from the design to prepress environment, U1ere
`are often problems, such as missing fonts, usually due to a lack of coordination
`between the two environments. One wants to catch these problems early and not
`discover them after they have taken up time, capacity and materials on the output
`device. It is also desirable to tix them locally. without having go back to the
`designer. Besides, they may not be design problems in the tirst place.
`
`Some of these problems can be handled with a preprocessing step. in this
`case preflight, which analyzes and assesses the PDL file before it is committed to
`output. Preflight either fixes errors in the file or alerts the operator in the prepress
`shop, who can go back and tix things at the application or else make sure the
`appropriate downstream resources are available.
`
`341
`
`
`
`Another reason for preprocessing is performing needed operations, such
`as trapping and imposition, which the RIP may not be capable of doing. Luminous
`Trap Wise and Press Wise are examples. They take in a PDL file, or multiple mes
`in the case of imposition, and output a single file that describes a trapped page or
`signature and that can then be sent to the RIP.
`
`An OPI (Open Prepress Interface) server is another example of a
`preprocessor. It takes in a PDL with the appropriate structuring conventions and a
`low resolution scanned image used for placement and layout, and replaces the
`latter with the high resolution version. This minimizes the size of the PDL file at
`the source and saves time in its transmission, always useful but especially so when
`the transfer occurs between two different sites over a low bandwidth link. It also
`takes advantage of existing prepress skills and experience in preparing images for
`printing, thus avoiding the need to replicate those skills at the design stage or in
`the application.
`
`Some preprocessing operations are candidates for integration with the
`RIP. Adobe has recently announced trapping and OPI as PostSLTipt Level 3 RIP
`functions.
`
`In-RIP Processing
`
`The next place to add capabilities and features is within the RIP. The
`RIP's essential function is to convert the POL representation of the job or
`document into a raster image description. A RIP is typically split and implemented
`in two pieces: an interpreter, which converts the PDL description into a display
`list, and an imager, which converts the display list into a raster image. The display
`list is an intermediate representation, more suited for conversion, usually with
`hardware assist, to a raster image than the PDL operators.
`
`The display list offers another interface for inserting value-add operations
`or "filters" into the digital data ftow. Several vendors have described a display(cid:173)
`list-based interface within the RIP: Linotype Hell's Delta List, HyWay Ferranti's
`GList, Harlequin's Display List Technology, and DaiNippon Screen's Page
`Object Module. A published display list interface offers third party vendors the
`opportunity to insert solutions. Among the display list filters that have been
`described or suggested are ones for trapping, vignette detection, OPI image
`replacement and page editing.
`
`Post-RIP Processing
`
`Another way to add capabilities to the system is with a postprocessor after
`the RIP. The postprocessor operates on a raster image in a specified image format.
`
`Examples of postprocessing operations are trapping (again), imposition.
`spatial resolution conversions and color calibration. While the RIP can do
`
`342
`
`
`
`screening (or halftoning) and output a binary, halftoned image, screening is often a
`postprocessing operation in high-end applications and is implemented by the
`modulator which drives the output device. Placing the screening operation at the
`end means that the image format is continuous tone (contone).
`
`Ute image format is key to enabling postprocessor operations. 111e format
`could be a simple page-oriented raster one, such as 11FF, or a more sophisticated
`fonnat such as Scitex CT/L W or 11FF/IT. At this conference two year ago.
`Venable et al. ( 1995) described Structured Images as a resolution and device(cid:173)
`independent compound raster image format. Earlier this year, Agfa announced
`Intcllipac, a compressible image format, as part of its Intellistream output
`manager for processing, transferring, and storing files.
`
`Improving Throughput
`
`Figures 4 and 5 show two system configurations enabled by pre- and post(cid:173)
`RIP processing in a graphic arts system. Figure 4 shows multiple RIPs for
`increased throughput on a single output device. ·rrus approach avoids the
`bottlenecks created by a single RIP or by a single page in a job. The preprocessor
`distributes the pages in a job across parallel RIPs. A postprocessor then collects
`and re-assembles the images output by the RIP and sends the completed job to the
`output device.
`
`Figure 4: Multiple RIPs, single output
`
`The multiple RIPs and single output in Figure 4 is a general description of
`Adobe's PostScript Extreme architecture. The page-parallel RIPping is enabled by
`Adobe's Portable Document Format (PDF). PDF is a derivative of PostScript.
`PDF has fewer operators than PostScript and is more self-contained, relying less
`
`343
`
`
`
`on external resources than PostScript does. PDF also has page independence and
`other document-oriented features not found in PostScript. llle simpler, more rigid
`structure of PDF means it is less complex and less flexible than PostScript, which
`leads to more predictable and greater throughput.
`
`output
`
`Output
`
`OUtput
`
`Figure 5: Single RIP, multiple outputs
`
`Figure 5 shows another approach to increased and more predictable
`throughput, this time in the context of a single RIP, driving multiple output
`devices. The RIP may be off-line and not directly connected or dedicated to any
`one output device; hence, Figure 5 also shows a storage element. As a result, the
`output image can become an "asset" to be managed in the workflow. 111e output
`devices can be identical, in which case there is load balancing; distributed,
`supporting the distribute-and-print model; or truly different, in which case they all
`use the same RIP. In other words, the file is RIPped once for output on multiple
`devices.
`
`A single RIP and the output image it produces provide a common
`reference point for multiple output devices. lllis eliminates the variability that can
`occur between the outputs of different RIPs. as not all RIPs interpret a PostScript
`file the same way. It also enables rework and reprinting without re-RIPping, which
`can be important in workflows where the application or PDL file is not available
`or easily accessible, where the goal is simply to save time, or where the proofer
`uses the same image file as the production output device.
`
`Postprocessing plays an important role in the approach shown in Figure 5.
`It prepares or converts the image file for an output device that may have a
`different resolution or color calibration than the device that the RIP targeted and
`that the image was originally prepared for. In effect postprocessing enables late
`binding.
`
`344
`
`
`
`Image Formats
`
`In systems, such as the one shown in Figure 5, that store, process,
`transmit and distribute raster images, the raster image format can have a
`significant effect on the postprocessing capabilities, and consequently on the
`workflow. In this respect, it is worthwhile briefly describing TIFF/IT and also
`llFF-FX, a new Tifr-ba..<;ed image format proposed for Internet Fax. What they
`have in common is both are image exchange formats for representing pages that
`contain a mixture of text, graphics and pictures.
`
`TIFF/IT (ISO, 1997) has its origins in the output of color electronic
`prepress systems or CEPS, which are raster oriented. It was standardized by ISO
`TC 130/WG2 (Graphic Technology!Prepress Data Exchange). TIFF/IT is
`intended for the digital distribution of color ads, which may contain a combination
`of pictures, graphics and text, and which may or may not be trapped. Commercial
`RIPs are available today that output TifHIT images.
`
`A TlfHIT file can contain multiple components: contone (CT) images,
`line art (L W) images, and high resolution contonc images (HC). Each is encoded
`differently: the CT component is uncompressed at a lower resolution than the other
`two components; the L W component has the highest resolution and is mapped to
`256 values and run-length encoded; the HC component is effectively CT with the
`resolution of L W. The primary color space for TIFF'/IT is CMYK. although
`CIELu'v' is among the available optiOrL'>.
`
`To combine overlapping multiple components into the final image,
`'IlFF/IT uses a simple imaging model: LW takes precedence over CT, except
`when the LW value indicates transparency, in which ca..<;e the underlying CT
`component or the paper is visible. Details are given in the 11FF/IT specification.
`
`The International Telecommunication Union (ITU), which sets
`telecommunications (including fax) standards, recently dealt with a similar
`situation. Separate ITU standards already exist for compressing and transmitting
`black-and-white text and line art, color graphics, and color pictures. However, for
`a page that combines all three, a fax system would be forced to pick a
`compression method, spatial resolution and color representation that would be a
`good choice for one of the three, but a poor choice for the other two.
`
`The ITU solution is the Mixed Raster Content (MRC) model, scheduled
`for final approval in October 1997 (ITU, 1997). Rather than describe new
`compression methods, MRC describes a 3-layer model for representing the
`scanned image data from a page that contains text, graphics and images, in either
`black-and-white or color. This way, each kind of content can be represented in
`way most suitable to it, independently of the other two and using one of the
`existing ITU coding methods.
`
`345
`
`
`
`In March 1997, Adobe Systems and Xerox (Mcintyre and Zilles, 1997)
`submitted a joint proposal for an Internet Fax file format called TIFF-r:x (r:x
`stands for "fax extended") to the IElF Internet Fax Working Group. lllis
`proposal, based on the current TIFF specification, adds the fields and values
`needed to represent the data produced by the suite ofiTU Recommendations for
`black-and-white and color facsimile. In particular, the proposal enables the nu
`Mixed Raster Content model in 'ITFF.
`
`The example in Figure 6 illustrates the MRC model applied to a page with
`colored text and a picture. The 3 layers of the MRC model are Foreground and
`Background, which are both multi-level, and the Mask, which is binary. Pixels in
`the Mask layer act as a switch. selecting pixels from the other two layers for the
`final image. When the Mask layer pixel value is 1, the corresponding pixel from
`the Foreground layer is selected; when it is 0, the corresponding pixel from the
`Background layer is selected. Details are given in the Information and Background
`section of the MRC Draft Recommendation (ITU. 1997).
`
`Foreground
`(Multilevel text color)
`e.g. JPEG, JBIG
`12bpp,100dpi
`
`:Mask
`(Bilevel text shape)
`:e.g. MMR
`:
`1 bpp, 400 dpi
`
`Background
`(Multilevel contone image)
`e.g. JPEG
`24 bpp, 200 dpi
`Figure 6: rru Mixed Raster Content (MRC) Model
`
`Each layer is coded independently of the other two, with a different choice
`of compression. color representation and spatial resolution. 111e choices shown as
`examples in Figure 6 are typical of a fax application. lhe Mask layer is a binary
`image, typically at high resolution and losslessly compressed. lhe Foreground and
`Background layers are usually at lower resolutions, which are integer factors of
`the Mask resolution (see Figure 6). The MRC model allows lossless compression
`
`346
`
`
`
`for the Mask layer and lossy or lossless compression for the other two layers. The
`primary color model for the MRC model in fax applications is CIELAB.
`
`For the colored text shown in Figure 6, the filled text shapes are in the
`Mask layer and the text color is in the Foreground layer; the corresponding part of
`the Background layer is blank. Wherever, the Mask layer pixels are black (value
`1), the Foreground color pixels are used for the final image. An alternative for this
`example is to place the colored text in the Foreground layer, and use a black
`rectangle in the Mask Layer to select it. Either way, the Mask layer would be
`coded using one of the ITU standard lossless binary compression methods, such as
`MMR (Modified Modified Read), and the Foreground layer would be coded using
`the ITU standard for lossy (JPEG) or lossless (JBIG) color compression.
`
`In this example, the picture is placed in the Background layer; the
`corresponding Mask layer pixels are white (value 0). so that the Background is
`selected for the tina! image. The Background layer could be coded using the ITU
`standard for lossy or lossless color compression.
`
`'The example in Figure 6 illustrates the use of all three layers in the MRC
`model. Other pages may use fewer layers. For example, a page containing a
`picture would use only the Background layer. A page of black-and-white text
`would use only the Mask layer, with the Foreground layer defaulted to black and
`the Background, to white. A page with text and a color bar chart could be
`represented by a combination of the Mask and Foreground layers.
`
`TIFF/IT and the TIFF-FX proposal are different solutions to representing
`mixed raster content documents. Inc ditierences are due to their different origins,
`requirements and applications. However, there is an overlap from which both
`might benefit when future extensions are considered. With this in mind, an
`informational package on the ITU MRC model and the 'OFF-FX proposal was
`requested and distributed at the TC 130/WG2 meeting in Chicago in May 1997.
`
`Conclusions
`
`Jl1is paper has described the progression from the basic framework for
`digital data operations shown in Figure 1 to the more elaborate framework shown
`in Figure 3. This progression has led to a proliferation of processing opportunities.
`a<> there are more places to insert functionality and change content. both early hut
`especially later in the flow.
`
`The growth in processing opportunities in turn has led to growth in
`commercial opportunities (more elements in the system and more pn.xluct choices
`for each element), standardization activities (more formats to keep track of, plug
`into and exchange between elements supplied by different vendors) and workJ1ow
`scenarios (more ways to combine. arrange and connect elements; more than one
`place in the flow to perform a function).
`
`347
`
`
`
`All this has significant consequences for both the system designer and
`system manager. Not all commercial prepress systems today would have all the
`elements or need all the functions shown in Figure 3. The system designer would
`choose elements and, for those elements, the product offerings appropriate to the
`work process, job requirements and market served.
`
`Given the multiple places where the file and its content can be touched and
`modified, another consideration is the choice of file to archive for retrieval and
`reprinting later: the application file, the PDL or the raster image output by the
`RIP. Again, the choice depends on the work process and job requirements, and
`would be made on a case-by-case basis.
`
`lne digital data operations in Figure 3 form the framework for developing
`and describing different workflow scenarios. Coordinating the possibilities tor the
`digital data operations calls for workflow management that can accommodate
`different elements, me formats, databases, platforms, networks, media, job
`requirements and user goals.
`
`Acknowledgments
`
`'This paper has benefited greatly from discussions with many of my Xerox
`colleagues, who helpful comments and inputs I would like to acknowledge.
`
`Literature Cited
`
`ISO, International Standards Organization
`1997
`ISO DIS 12639, Tag Image File Format for Image Technology
`(TIFF/IT)
`
`ITU, International Telecommunication Union
`1997
`Draft Recommendation T.44, Mixed Raster Content (MRC)
`
`Mcintyre, L and Zilles, S.
`1997
`''File Format for Internet Fax," IETF Internet Draft available at
`http://www .ietf.org/html.charters/fax -charter .h trnl
`
`Venable, D, Buckley, R. and Yamada, T.
`1995
`''Managing and Representing Workflow in Prepress Applications,"
`TAGA Proceedings, pp. 373-385
`
`The product names mentioned in this paper are or may be the trademarks or registered
`trademarks of their respective owners.
`
`348
`
`