GB2348773A

GB2348773A - User interface

Info

Publication number: GB2348773A
Application number: GB0016919A
Authority: GB
Inventors: Richard A Holub
Original assignee: Individual
Current assignee: Individual
Priority date: 1996-02-26
Filing date: 1997-02-25
Publication date: 2000-10-11
Anticipated expiration: 2017-02-25
Also published as: GB2348700B; GB0016921D0; GB2348773B; GB2348700A; GB0016919D0

Abstract

A user interface for controlling colour reproduction at multiple sites, which is implemented on a computer using object-orientated components, has an abstract menu object comprising a command name object and a submenu command object derived from the abstract menu object, and a menu command object derived from the abstract menu object comprising a plurality of submenu command objects and a menu bar object derived from the abstract menu bar comprising a plurality of the menu command objects and a select menu.

Description

2348773 A SYSTEM FOR DISTRIBUTING AND CONTROLLING COLOR REPRODUCTION AT

MULTIPLE SITES

FIELD OF THE INVENTION

This invention relates to a system (method and apparatus) for distributing and controlling color reproduction at multiple sites, and particularly to, a system for distributing and controlling the color output of rendering devices, such as proofing devices and presses. at multiple sites or nodes of a network to provide a uniform appearance of color within the output colors attainable at each rendering device. The system is controlled by computers at each node and utilizes a data structure, referred to herein as a Virtual Proof. to store and distribute color transformation information in the network. Color image data representing one or more pages or page constituents can be distributed to the nodes separately from the Virtual Proof BACKGROUND OF THE INVENTION

In the printing and publishing industry, the increasing modularity of manufadturing operations is enabling custornization of products. At the same time, pressures to reduce inventories and to keep them fresh aredriving a trend toward just-in-time production and stocking. Wherever the manufacturing can be decentralized and distributed geographically. justin-time production is facilitated because producers are closer to consumers in space and time. There is an ecological dividend, as well, in reduced demands on the transportation system. Overall product cost may decrease with shipping expense. At the same time. however, the challenge of maintaining uniform quality across a network of production sites increases. Minimizing startup waste gains in importance as does compensating for uneven skill and experience of operators. Color is a key variable to control because it affects product appearance and perceived quality.

Today for example. a magazine with a national circulation of 3 million may be printed at 5 regional plants scattered across the nation. Distribution (transportation and 2 postage) generally account for one third of the cost of the product while transit time has a significant impact on product "freshness," i.e., the timeliness of the information delivered.

Production is as centralized as it is partly in order to maintain reasonably homogeneous quality. Nevertheless, printed color varies within a press run and from site to site because therehave been only limited means of coordinating control of product appearance among sites. The scope and significance of this problem is apparent when one considers how much commerce and economic activity are leveraged by advertising and that generally more than 60% of all printing is advertising-related. Analogous problems also arise in other media, particularly now that digital video images can be edited in real time and broadcast directly.

The preceding paragraphs have spoken about parallel mass-production at multiple sites. Publishing is also distributed in the sense that the sequential steps of preparation for volume production occur at distinct sites. as illustrated in Figure 1. Oftentimes, the sites represent different business entities (for example, an advertising agency, a publisher. or an "engraver") which are geographically separated. Solid lines in Figure I represent links cormecting the sites in the production process. Overlaid in Figure I are dotted boundaries indicating a cluster of pre-publishing facilities which handle sequential phases of the process under Product Prototype 1, and printing facilities which may be involved in parallel Volume Production 2.

Currently prevalent volume printing technologies such as offset lithography. employ a printing "plate" which bears fixed information and is the tool or die of volume production. The tool is mounted on a press and numerous copies of the printed product are stamped out. For technologies such as ink jet and electrophotography the information on the plate can be changed from one revolution of the press to the next. This technological development enables significant product customization and is compatible with Just-in-time production scenarios. It also enables process control in which the electronic data flowing to the device are modified to adapt to changes in the marking engine. However. the consistency (or repeatability) of these processes makes them even more susceptible to regional variations in quality across the production sites than lithography and its relatives.

1 3 For all of the printing technologies mentioned, there is a common problem of uniform and accurate color reproduction. Analogous problems also exist in other media for distributing color graphic or image content, such as CDROM or the Internet. Consider an advertiser in New York, physically removed from the five production sites mentioned above, or the more numerous sites that may be involved in the future. There is a keen interest in having the product portrayed in as faithful an accord with the advertiser's artistic conceptions as possible, even when the ad is to appear in different publications printed on different substrates by different machinery or in the same publication disseminated through different media.

Today, the approval cycle, the means by which print buyer and printer reach contractual agreement-about the acceptability of product, often proceeds as outlined in Figure 2. in the publication segment of the industry. Phases or functions of production are enclosed in ellipses la, lb and le and key products of theses functions are enclosed by rectangles 3, 5, 6, 7. 8 and 9. The dashed line between creation I a and prepress I b shows the blurring of those functions in the development of intermediate products, such as page constituents like lineart, images, text and com s. Prepress lb on the way to film 5 may p include rasterization, separation and screening 4. However. acceptance of computer toplate technology will blur the boundary between prepress lb and production Ic.

The long, heavy boundary line between press-proofing in low volume reproduction 1c and high volume production 2 represent the distinctness of the two functions: the former is carried out by engravers or commercial printers. Note that volume production 2 ma) occur at multiple sites. Linkages in the approval process are shown by arcs 10a and 10b at the bonom of Figure 2, where 10a is the traditional off-press proof and 10b is a press proof. The transactions in the approval process involve one or more generations of static proofs which are prepared with limited regard for the capabilities of the final. volumeproduction devices. In other words. there is no feedback from production to earlier functions. The process results in idle time -for -equipment and personnel and waste of consumables (paper, ink etc.) Furthermore. it usually does not give the pnnt buyer an) direct say about the appearance of the ultimate product unless the buyer travels to the printing plant, an expensive proposition.

C -4- The workflow for commercial printing is slightly different from that described above, since press-proofs are seldom used and the print buyer or his agent often go to the printer's for approval. However, the essential lack of feedback is also prevalent in the commercial environment as well.

It is clear that a common language of color could insure improved communication. control and quality throughout the sites of Figure 1. The common language is a color space, typically based on the internationally accepted Standard Observer which quantifies color in terms of what normal humans see. rather than in terms of specific samples or instances of color produced by particular equipment. The Standard Observer is the basis of device- independent, colorimetric methods of image reproduction and is defined by the Commission Internationale de L'Eclairaize in CIE Publication 15.2. 1986. Central Bureau of the CIE. Box 169, Vienna, Austria. Approximately uniform perceptual color spaces based upon the Standard Observer are also discussed in this publication.

Color Space is defined as a three-dimensional, numerical scheme in which each and every humanly perceivable color has a unique coordinate. For example, CIELAB is a color space defined by the CIE in 1976 to simulate various aspects of human visual performance. Color in the next section will refer to CIE color or what we see, while colorant will refer to particular physical agents. such as dyes, pigments, phosphors, and the like that are instrumental in producing sensations and perceptions of color in a human at rendering devices. such as presses and video screens.

Prior Art

An early machine for converting color image data to colorant specifications for a 3 or 4-channel reflection reproduction process was described by Hardy and Wurzburg (Color correction in color printing, J. Opt. Soc. Amer. 38: 300-307, 1948.) They described an electronic network provided with feedback to control convergence to the solution of an inverse model of colorant mixture and produce 4-colorant reproductions "indistinguishable" from -3-colorant reproductions made under like conditions. The set point for the control was the color of the original. This work serves as a starting point for many subsequent developments in the art particularly as regards device independent color reproduction

1 technologies and "color separation," i.e., the preparation of printing plates for 3 or more colorants.

In U.S. Patent No. 7,790,844, Neugebauer discloses a system to extend the HardyWurzburg machine. It describes the capture and representation of color imagery in a colorimetric (or device independent) coordinate system. To enable an operator to judge the effect of color corrections while he is makin g these color corrections, the system provides for a soft proof realized by projecting video images onto the type of paper stock to be used in the final reproduction with careful regard to making the surround illumination and viewing conditions comparable to those prevailing when the final product is viewed. The objective of the soft proof was to simulate a hard copy proof or final ptint. This is in contrast to U.S. Patent-No. 4.500.919, issued to Schreiber, which discloses a system to match the hard copy to the monitor image.

Concerning models of color formation by combination of colorants, Pobboravsky (A proposed engineering approach to color reproduction, TAGA Proceedings, 1962, pp. 127165) first demonstrated the use of regression techniques ("curve fining") to define mathematical relationships between device independent color ("in the CIE sense") and amounts of colorant with accurate results-. _T he mathematical relationships took the form of low order polynomials in several variables.

Schwartz et al. (Measurements of Gray Component Reduction in neutrals and saturated colors, TAGA Proceedings, 19 85, pp. 16-27) described a strategy for inverting "forward" models (mathematical functions for converting colorant mixtures to color.) The algorithm was similar to Hardy and Wurzburg's but implemented with digital computers: it consists of iteratively computing (or measuring) the color of a mixture of coiorants. comparing the color to that desired and modifying the colorants in directions dictated by the gradients of colorants with respect to color error until color error is satisfactorily small. Color error is computed in CIE uniform -coordinates. The context of the work was an implementation of an aspect of the art known as Gray Component Replacement (GCR.) Because normal human color perception is inherently 3-dimensional. the use of more than 3) colorants is likely to involve at least one colorant whose effects can be simulated by a mixture of two or more of the others ("primaries.") For example. various amounts of black ink can be matched by specific mixtures of the "primary" subtractive colorants cyan, magenta and yellow. The goal of Schwartz et al. was a method for finding colorimetrically -,;jivalent ("indistinguishable" in Hardy and Wurzburg's words) 4-colorant solutions to the problem of printing a given color that used varying amounts of black. Additional colorants (more than 3) are used to expand the gamut; black enables achievement of densities in reflection reproduction processes that are not otherwise available. A gamut is the subset of human perceivable colors that may be outputted by a rendering device. However, increased reliance on black through GCR has other important dividends: a) there is an economic benefit to the printer and an environmental benefit at large in using less colored ink. b) use of more black affords better control of the process.

Boll reported work on separating color for more than four colorants (A color to colorant transformation for a seven ink process, SPIE Vol. 2170, pp. 108-118, 1994. The Society for Photo-Optical and Instrumentation Engineers, Bellingham, WA.). He describes the "Supergamut" for all seven colorants as a union of subgamuts formed by combinations of subsets of 4at-a-time of the colorants. Because of the manner in which his subsets are modelled, the method severely limits flexibility in performing GCR.

Descriptions of gamuts in colorimetric terms date at least to Neugebauer (The colorimetric effect of the selection of printing inks and photographic filters on the quality of multicolor reproductions, TAGA Proceeding, 1 -.'56, pp. 15-28.) The first descriptions in the coordinates of one of the CIE's uniform color spaces are due to Gordon et al. (On the rendition of unprintable colors, TAGA Proceedings, 1987, pp. 186-195.) who extended the work to the first analysis of explicit gamut operators - i. e.. functions which map colors from an input gamut to correspondents in an output gamut.

A detailed review of requirements of and strategies for color systems calibration and control was published by Holub. et al. (Color systems calibration for Graphic Arts. Parts I and 11, Input and output devices, J. Imag. Technol., 14: 47-60. 1988.) These papers cover four areas: a) the application of color measurement instrumentation to the calibration of devices, b) requirements for colorimetrically accurate image capture (imaging colorimetry,) c) development of rendering transformations for 4colorant devices and d) requirements for soft proofing.

1 Concerning the first area (a), U.S. Patent No. 5.272,518, issued to Vincent, discloses a portable spectral colorimeter for performing system-wide calibrations. The main departure the work of Holub et al., just cited, is in the specification of a relatively low cost design based on a linearly variable spectral filter interposed between the object of measurement and a linear sensor array. Vincent also mentions applicability to insuring consistent color across a network, but does not discuss how distributed calibration would be implemented. There is no provision for self-checking of calibration by Vincent's instrument nor provision for verification of calibration in its application.

U.S. Patent No. 5,107,332. issued to Chan, and U. S. Patent No. 5,185,673, issued to Sobol. disclose similar systems for performing - closed- loop control of digital printers. Both Chan and Sobol share the following features: 1) They are oriented toward relatively low quality, "desktop" devices, such as ink jet printers. 2) An important component in each system is a scanner, in particular, a flat-bed image digitizer. 3)) The scanner and printing assembly are used as part of a closed system of calibration. A standardized calibration form made by the printing system is scanned and "distortions" or deviations from the expected color values are used to generate correction coefficients used to improve renderings. Colorimetric calibration of the scanner or print path to a device independent criterion in support of sharing of color data or proofing on external devices was not an objective. 4) No requirements are placed upon the spectral sensitivities of the scanners RGB channel sensitivities. This has ramifications for the viability of the method for sets of rendering colorants other than those -used -in the closed printing system, as explained below.

In Sobol, the color reproduction of the device is not modelled; rather the "distortions" are measured and used to drive compensatory changes in the actual image data, prior to rendering. In Chan, there appears to be a model of the device which is modified by feedback to control rendering. However. colorimetric calibration for the purposes of building gamut descriptions in support of proofing relationships among devices is not disclosed.

Pertaining to item (b) of the Holub. et al. paper in J. Imaging Technolog and to the foregoing patents. two articles are significant: 1) Gordon and Holub (On the use of 0 - 8 linear transformations for scanner calibration, Color Research and Application, 18: 218219, 1993) and 2) Holub (Colorimetric aspects of image capture, IS&T's 48th Annual Conference Proceedings, The Society for Imaging Science and Technology, Arlington, VA. May, 1995, pp. 449451.) Taken together, these articles demonstrate that, except when the spectral sensitivities of the sensor's channels are linear combinations of the spectral sensitivity functions of the three human receptors, the gamut of an artificial sensor will not be identical to that of a normal human. In other words, the artificial sensor will be unable to distinguish colors that a human can distinguish. Another consequence is that there is generally no exact or simple mathematical transformation for mapping from sensor responses to human responses, as there is when the linearity criterion set forth in this paragraph is satisfied by the artificial sensor.

To summarize the preceding paragraphs: The objective of measuring the colors of reproduction for the purpose of controlling them to a human perceptual criterion across a network of devices in which proofing and the negotiation of approval are goals is best served when the image sensors are linear in the manner noted above.

Results of a colorimetric calibration of several printing presses were reported Oy Holub and Kearsley (Color to colorant conversions in a colorimetric separation system. SPIE vol. 1184. Neugebauer Memorial Seminar on Color Reproduction, pp. 24-33. 1989.) The purpose of the procedure was to enable workers upstream in the production process in a particular plant to be able to view images on video display devices, which images appeared substantially as they would in production, consistent with the goals of Neugebauer in U.S. Patent No. 2,790,844. Productivity was enhanced when design could be performed with awareness of the limitations of the production equipment. The problem was that the production equipment changed with time (even within a production cycle) so that static calibration proved inadequate.

In U.S. Patent No. 5.182,721. Kipphan et al. disclose a system for taking printed sheets and scanning specialized "color bars" at the margin of the sheets with a spectral colorimeter. Readings in CIELAB are compared to aim values and the color errors so generated converted into corrections in ink density specifications. The correction signals are passed to the ink preset control panel and processed by the circuits which control the

1 inking keys of the offset press. Operator override is possible and is necessary when the colorimeter goes out of calibration, since it is not capable of calibration self-check. Although the unit generates dam useable for statistical process control, the operator must be pro-active in sampling the press run with sufficient regularity and awareness of printed sheet count in order to exploit the capability. The process is closed loop, but off-line and does not read image area of the printed page. Important information regarding color deviations within the image area of the press sheet is lost by focussing on the color bars.

On page 5 of a periodical "Komori World News" are capsule descriptions of the Pnnt Density Control System, which resembles the subject of Kipphan et al. Also described is the Print Quality Assessment System, which poises cameras over the press. The latter is primarily"oriented toward defect inspection and not toward on-line color monitoring and control.

Sodergard et al. and others (On-line control of the colour print quality guided by the digital page description, proceedings of the 22nd International Conference of Printing Research Institutes, Munich, Germany, 1993 and A system for inspecting colour printing quality, TAGA Proceedings, 1995) describe a system for grabbing frames from the image area on a moving web for the purposes of controlling color, controlling registration and detecting defects. The application is in newspaper publishing. Stroboscopic illumination is employed to "freeze" frames of small areas of the printed page which are imaged with a CCD camera. The drawback of the Sodergard et al. system is that color control lacks the necessary precision for high quality color reproduction.

Optical low pass filtering (descreening) technology relevant to the design of area sensors for imaging colorimetry is discussed in U.S. Patent No. 4,987,496. issued to Greivenkamp, and Color dependent optical prefilter for the suppression of ahasing artifacts. Applied Optics. 29: 676-684. 1990.) Paul Shnitser (Spectrally adaptive acousto-optic tunable filter for fast imaging colorimeTry. Abstract of Successful Phase I Proposal to U.S. Dept. of Commerce Small Business Innovation Research Program. 1995) and Clifford Hoyt (Toward hip-her res. lower cost quality color and multispectral imaging. Advanced Imaging, April '95) have discussed the applicability of electronically tunable opticalispectral filters to colorimetric imagin!2.

10- In "Thin-film measurements using SpectraCube"m," (Application Note for Thin Film Measurements, SD Spectral Diagnostics Inc., Agoura Hills, CA 91301-4526) Garini describes a spectral imaging system employing "...a proprietary optical method based on proven Fourier spectroscopy, which enables the measurement of the complete visible light spectrum at each pixel..."

The applicability of neural network (and other highly parallel and hybrid) technologies to the calibration and control of rendering devices has been considered by Holub ("The future of parallel, analog and neural computing architectures in the Graphic Arts," TAGA Proceedings, 1988, pp. 80-112) and U.S. Patent No. 5,200,816, issued to Rose, concerning color conversion by neural nets.

A formalism from finite element analysis is described in Gallagher, "Finite element analysis: Fundamentals." 1975. Englewood Cliffs, N.J.: Prentice Hall, pp. 229-240 for use in the rapid evaluation of color transformations by interpolation.

Area (d) of the earlier discussion of Holub et al.'s review referred to principles guiding the design and application of softproofing: methods of calibrating video displays. evaluation of and compensation for illumination and viewing conditions, representatign of how imagery will look on client devices and psychophysical considerations of matching appearance across media.

In the article "A general teieproofing system," (TAGA Proceedings, 1991. The Technical Association of the Graphic Arts. Rochester, NY) Sodergard et al. and others discuss a method for digitizing the analog image of an arbitrary monitor for transmission through an ISDN telecommunications link to a remote video display. The method involves the transmission of the actual image data, albeit at the relatively low resolution afforded by the frame buffers typical of most displays. This method lacks any provision for calibration the or verification of the devices at either end of a link and also lack4 data structures needed to supp, - remote proofing and negotiation of color approval.

In U.S. Patent No. 5.231.481. Eouzan et al. disclose a svstem for controlling a projection video display based on cathode ray tube technology. A camera is used for capturing image area of a display. The procedures are suited to the environment in which the displays are manufactured and not to where they are used. Concepts of colorimetric 1 calibration of the display and control of display output to a colorimetric criterion are not disclosed.

In U.S. Patent No. 5,309,257, Bonino et al. disclose a method for harmonizing the output of color devices, primarily video display monitors. In a first step, measurements of the voltage in vs. luminance out relationship are made for each of the three display channels separately and then the V/L functions of all the devices are adjusted to have a commonly achievable maximum. This is assumed to insure that all devices are operating within the same gamut - an assumption which is only true if the chromaticities of the primaries in all devices are substantially the same. The single-channel luminance meter (a "photometer") described as part of the preferred embodiment does not permit verification of the assumption. Bonino et al. thus employs photometric characterization of devices and lacks a colorimetric characterization.

The Metric Color Tag (MCT) Specification (Rev 1. 1 d, 1993, Electronics for Imaging, Inc., San Mateo, CA is a definition of data required in data files to al1ow color management systems to apply accurate color transformations. The MCT thus does not provide a file format defining the full specification of color transformations in the context of distributed production and color-critical remote proofing.

In contrast to the MCT, the International Color Consortium (ICC) Profile Format is a file format, and is described in the paper, International Color Consortium Profile Format (version 3.01. May 8, 1995). A "profile" is a data table which is used for color conversion - the translation of color image data from one color or colorant coordinate system to another. The ICC Profile Format provides for embedding profiles with image data. This generates large data transfers over a network whenever profiles are updated. Further, the ICC Profile, Representation of devices in the ICC Profile Format is limited in supporting L_ scnr" (scanner), "mntr" (video display monitor) and "prtr" (printer) device types. and is not readily extendable to other types of devices.

Interactive remote viewing is described for "imagexpo" application software from Group Logic, Inc., in the article "Introducing imagexpo 1. 2: Interactive remote viewing and annotation software for the graphic arts professional" and "Before your very eyes."

(reprinted from Publishing & Production Executive. Aug. '95). which acknowledges that extant tools do not enable remote handling of colorcritical aspects of proofing.

SUMMARY OF THE INVENTION

A general object of the present invention is to provide a system for controlling and distributing color reproduction in a network of nodes having rendering devices or systems, such as volume production machinery, pre-press and proofing devices, in which colors ha V P_ reproduced at each rendering deviceha-4ft substantially the same appearance within the output colors attainable by the rendering devices.

Also an object of the present invention is to provide a system for controlling and distributing color reproduction in a network with a dynamic virtual proof data structure which includes information needed for color reproducing at rendering devices in the network.

Another object of the present invention is to provide a system for controlling and distributing 4a4 color reproduction which uses color instrumentation at rendering devices in calibrating the color of the devices and in subsequent verification of such calibration.

A further object of the present invention is to provide a system for controlling and distributing color reproduction which communicates color accurately between rendering devices at multiple sites through virtual proofing. thereby providing color quality control.

Still a further object of the present invention is to provide a system for controlling and distributing color reproduction to control and coordinate multiple site parallel. volume production, and to control color quality of reproduction at those sites by interfacing with feedback control for production machinery.

Yet another object of the invention is to provide a system for distributing and controlling color reproduction which provides forrevising color for remote printing or display without requiring retransmission of large image data files.

A further object of the present invention is also to provide a systern for distributing and controlling color reproduction having an improved user interface and allows for an arbitrary number of colorants. which may exceed four.

1 13 - in addition, an object of the present invention is to provide a system for distributing and controlling color reproduction in a network of nodes which provides each node with information needed for rendering transformations of color image data in response to both color calibration data and user color preferences.

In summary, the present invention integrates in a system improved instrumental embodiments which enable more precise control of color reproduction than prior art systems. especially when combined with improved characterization of devices and building color conversion transforms. The invention also incorporates improved data and file structures along with procedures and a user interface, which specifically address accurate color sensing and rendition in a networked environment.

Briefly described, a system embodying the present invention provides for controlling color reproduction of input color image data representing one or more pages (or page constituents) in a network having nodes (or sites). Each one of the nodes comprises at least one rendering device. The system distributes the input color image data from one of the nodes to other nodes, and provides a data structure (virtual proof) in the network. This data structure has components shared by the nodes and other components presept onl5at each node. Next. the system has means for providing color calibration data at each node characterizing output colors (colorants) of the rendering device of the node, and means for producing at each node, responsive to the color calibration data of the rendering device of the node. information for transforming the input color image data into output color image data at the rendering device of the node. The information is then stored in the data structure in different ones of the shared and other components. Means are provided in the systern for transforming at each node the input color image data into output color ima2e data for the rendering device of the node responsive to the information in the data structure. The rendering device of each node renders a color reproduction of the pages responsive to the output color image data, wherein colors displayed in the reproduction at the rendering device of each node appear substantially the same within the output colors attainable bv the rendering devices.

The system may further have means for verifying at each node that the informatioll for the rendering device of the node properly transformed the input color image data into C. -14- the output color image data, and means for revising the information stored in the data structure at the node responsive to results of the verifying means. Shared components of the data structure may also store color preferences selected by a user. The information producing means of the system may further operate responsive to both the color calibration dam and the color preferences. The rendering devices in the system can provide color reproductions having three or four output colors, and may provide more than four output colors.

The Virtual Proof feature of the system provides a file structure enabling accurate communication and control of color in a distributed production environment of the network. It contains data for procedures necessary to mediate conversions from the color of input image data to colorant, and vice versa- in such a way that the recording device at one particular node of the network can represent the output of other equipment to the best degree possible. It stores data representing information for transforming color at multiple nodes. The integrity of the data of the Virtual Proof is assured by continual calibration. verification and recalibration by color measurement instruments which are themselves the subject of ongoing calibration. The Virtual Proof embodies data on the extent and location of color error in product useful to automatic systems for the control of production rendering devices, such as high volume presses. Because it is a mechanism for interpreting color data to a device, the Virtual Proof provides reviseability of color data up to and during production.

The foregoing and other features, objects, and advantages of the invention will become more apparent from a reading of the following detailed description in connection with the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure I is a diagram of the typical sites involved in preparing volume production of color products.

Figure 2 is a diagram showing a conventional workflow for publication printing.

Figure -3)A shows the system in accordance with the present invention.

1 - Figure 3B shows a configuration of a color measuring instrument sensor for a video screen display.

Figure 3C shows geometries of a color measuring instrument sensor for making non-contact color measurements of reflective substrate.

Figure 3D shows use of a color measurement instrument to estimate a composite spectral function given knowledge of the underlying spectral functions of the colorants being mixed.

Figure 4A illustrates the devices of the system separated into device classes inheriting various structures and procedures for color calibration and transformation.

Figure 4B is a process diagram for color transformation of a class of devices including linear color measuring instruments.

Figure 4C is a process diagram for color transformation of a class of rendering devices including video-color displays.

Figure 5 is a process diagram for calibrating a class of rendering devices including printers and presses at a node in the sYstem of Figure 3A to provide color transformation information.

Figure 6A is a flow chart detailing step I in Figure 5, preparing lineariZation functions.

Figure 6B is a flow chart detailing step 2 in Figure 5. rendering calibration forms.

Figure 7 is a flow chart detailing step 3 in Figure 5. measuring calibration forms and providing calibration data.

Figure 8 is a flow chart detailing step 4 in Figure 5, building a forward model based on the calibration data from step 3 of Figure 5.

Figure 9A is a flow chart detailing step 5 in Figure 5, preparing gamut descriptor data for the rendering device and preparing a forward model table based on the forward model from step 4 in Figure 5.

Figure 9B is an illustration of the operators and operands evaluating the polynomial function of the forward model in the case of two independent (colorant) variables. C and M.

16- Figure 9C depicts a hypercube in the coordinates of the Cyan, Magenta. Yellow and Black colorants in which all colors producible with a CMYK printer are contained within the hypercube.

Figure 9D is an illustration of a data structure for interpolation in 33 dimensions which may use either pre- or post-conditioning look-up tables.

Figure 9E is a graphical illustration of linear interpolator in two dimensions.

Figure IOA is a flow chart detailing step 6 in Figure 5, inverting the forward model table to provide a prototype transformation table.

Figure IOB is an example of the hypercube of Figure 9C where the colorant values are transformed into device independent color coordinates.

Figure I I is a flow chart detailing step 7 of Figure 5. finishing the gamut descriptor data.

Figure 12 is a flow chart detailing step 8 of Figure 5. filling in any missing entries of prototype transformation table, computing black utilization (GCR) functions for all colors of the table having multiple black solutions and marking unprintable colors NULL.

Figure 13 is a flow chart detailing step 9 of Figure 5 which includes: converting. colorants of prototype transformation table based on black color data, building color to color' transform table based on gamut configuration data; and combining the color to color' transformation table and the converted prototype transformation table to provide a rendering table.

Figure 14 is an illustration of the construction of a simple gamut operator of the eamut configuration data embodying properties of invertibility and reciprocality.

Figures 15A and 15B show the constituents of local and shareable components in the data structure of the Virtual Proof.

Figure 15C is an example of a tagged file format for the shared components of the Virtual Proof of Figures 15A and 15B.

Figure 16A is a flow chart of the process for calibrating a rendering device having more than four colorants bv addine non-neutral auxiliary colorants to a rendering transformation.

1 1 17 - Figure 16B is a flow chart of the process for calibrating a rendering device having more than four colorants by adding a neutral auxiliary colorant to a rendering transformation: Figures 16A and 16B are shown connected.

Figure 17 is a flow chart showing the process for preparing a gamut filter. a data structure which facilitates the comparison of gamuts of two or more rendering devices.

Figures 18A and 18B are a flow chart for virtual proofing using the system of the present invention.

Figure 19 is a flow chart of the verification procedures employed to estimate process variations based on measurements of color errors.

Figure 20 is a flow chart of the process of preparing a three-dimensional color histogram of color image data in providing a resolution-independent analysis of color error relative to a criterion.

Figure 21A is a menu of the Graphical User Interface (GUI) to the application software to enable configuration of the network of nodes, remote conferencing and proofing and oversight of the processes involved in calibrating devices in the systems of Figure 3A.

Figure 21B is a menu at the second level of hierarchy in the GUI to provide access to tools for configuring the network connections and communications protocols.

Figure 21 C is a menu at the second level of hierarchy in the GUI to enable a User to manipulate the process of making color transformations at a rendering device.

Figure 2 1 D is a menu at the second level of hierarchy in the GUI to enable a user to oversee the process of using color transformations at a rendering device.

Figure 21E is a menu at the third level of hierarchy of the GUI which depicts the User interface to black utilization tools in providing black color data, and to neutral colorant definitions.

Figure 21F is a menu at the third level of hierarchy of the GUI which depicts the User interface to gamut processing at a rendering device in the network.

0 -18- DETAILED DESCRIPTION OF THE INVENTION

Referring to Figure 3A, the system 100 of the present invention is shown. ystern 100 has a r-1-vork 11 having a pipe Ila through which multiple nodes (or sites) of network 11 can be linked for dam flow between nodes. Network 11 may be a telecommunication network, WAN, LAN (with a server) or Internet based. Two types of nodes are present in system 100, prototype nodes 102 and production nodes 104. For purposes of illustration, only a general node of each type is shown in Figure 3A, however there may be multiple nodes of each type in network 11. Network 11 is modifiable to be configured by any one node to connect any two or more nodes in system 100. Each node has a micro-processor based computer, with a network communication device, such as a modem. which is part of a system having a rendering device for producing color reproduction and color measuring instrument (CMI) for measuring the color output of the rendering device. The computer may be a programmable general purpose computer or mainframe computer. Although a computer at a node is preferred, alternatively, the computer may be omitted at a node and the node operated remotely via pipe Ila from another node.

Prototype nodes 102 allow a user to perform pre-publishing functions in system 100, such as proofing (hard or soft), as well as the input of digital color image data. A user may interface with the node through standard interface devices. such as a keyboard or mouse. Rendering devices in system 100 define any type of system or device for presenting a color reproduction in response to digital color signals. The rendering devices of prototype node 102 are proofing devices, such as video screen display device 17 or proofer device 16. Proofing device 16 are hard copy devices. such as analog film-based devices, dye diffusion thermal transfer devices, ink jet printers. xerographic printers, and other similar devices. Video screen display 17 is useful for soft proofing (without a hard copy) and may be a high resolution video projection display for projecting images onto paper substrate(s) to be used in volume reproduction with resolution sufficient to reveal moird, i.e., halftone frequency beating patterns. Note that proofing devices are typically used to represent the performance of a client. such as a production rendering device described below at production node 104. The CMI associated with each proofing device is referred to as a standard observer meter (SOM) 13 and provides color measurement data from images from the proofing device.

SOMs 13 have the capability of measuring color as humans see it using the internationally accepted Standard Observer mentioned earlier, and will be described in more detail later.

One of the p-.;-- publishing functions supported by prototype node 102 is designing page layouts. Accordingly, a user or designer at nodes 102 can input digital color graphical/image data from a storage 19. which may be a hard drive, or from other sources. The color image data may consist of page layouts having images at low and high resolutions, such as RGB- A user at the node can define color preferences for rendering of the color image data. and later modify such preferences. The rendering of the inputted color image data at rendering devices to create soft or hard proofs is discussed later.

Production nodes 104 of network 11 control a production rendering device via the device's control system. -Production rendering devices include volume production machiner. such as press 15, which includes gravure presses, offset presses. electrophotographic printing machines, ink jet printers. flexographic presses, and the like. In addition, production nodes 104 may also have one or more rendering devices and SOMs 13 of a prototype node 102, such as proofing devices 20, which allows proofing to occur at a production site. CMIs of node 104 are called imagicals. Like SOMs 13 of prototy pe nodes 102, imagicals 14 provide color data for images rendered by press 15 in color coordinates of the Standard Observer. Proofing devices at prototype nodes 102 may also be outfitted with imagicals 14 which may incorporate a SOM 13. The differences between SOMs 13 and imagicals 14 will be described later. The main distinctions between production rendering devices and proofing devices are that the proofing devices are typically called upon to represent some other device. referred to herein as a client. such as printing press 15 of a production node 104 and such presses may have interfaces to and mechanisms of control that are different from those of proofers.

At a production rendering device.. the circuitry at node 104 differs from node 102 because it interfaces with inking control systems -of the production rendering device to maintain its color quality during volume reproduction. This allows control of the actual marking process. including variables such as ink film thickness. toner density. and the like. b supporting on-line cotorimetry from a CMI within image areas of printed sheets. Analysis of the CMI data can be used to produce error signals in CIELAB color difference units or in other units suitable for interface to commercially available inking control systems.

The nodes 102 and 104 each provide circuitry, which includes the computer and modem described above, for computation and communication. This circuitry operates responsive to programming of interface and application software stored at the nodes 102 and 104, and received user commands at the nodes. The processes defined by such programming operate system 100. The circuitry perform several functions. First, it accepts measurement data from CMIs and computes color transformation functions to translate between humanperceptible colors of the measurement data into rendering device colorant values. Second. it processes and transmits color graphical/image dam from one node or site in a network 11 to another. Third, it can issue reading instructions to CMIs mounted on a rendering device to measure rendered color images, and issue rendering instructions to a rendering device at the node using a stored color transformation. Fourth, the circuitry performs communications in system 100 in accordance with protocols for local or wide area networks, or telecommunications networks based on modem (either direct or mediated by Internet connection - note that Internet connectivity is not limited to modemj satellite link, TI or similar leased line technologies, ISDN, SMDS and related switched linkages, including Asynchronous Transfer Mode-enabled versions, TCP/IP, token ring. and the like. Fifth, the circuitry implements calibration of rendering devices to a common, human perceptible language of color. such as CIE, defined earlier, by producing and storing color transformation information. Sixth, the circuitry performs verification of the calibration of the rendering device to maintain accuracy of the stored color transformation information. These and other capabilities of the circuitry at a node will become apparent from the below discussion which describe further the processes referred to above.

A feature of the present invention is a data structure operating within system 100 called a Virtual Proof, hereafter called VP 12. The VP data structure is a file structure for storing and transmining files representing color transformation information between network 11 nodes. The contents of these files is outlined later. The VP is dynamic because it can be revised by nodes to assure the output color (colorants) of a rendering device using data from CMIs. Preferably. the VP does not contain color image data representing page layouts.

1 and is associated with the page layouts. However, it can alternatively have files storing some image data, although being separable from the often bulky high resolution image data representing the page layouts. The VP has components or files shared by the nodes in network 11, and local components or files present only at each node. Shared components are those useful by more than one node in network 11. while local components are particular to information of each individual node's rendering device. Shared components are transmitted by the circuity of each node to other nodes of network 11 via pipe Ila. Preferably, VP shared components are compact for transmission from node to node in network I I quickly. These shared VP components include files representing the user color preferences inputted at node 102 or 104, which is needed by each node in calibrating its rendering device. Each rendering device has its own version of a VP stored at its associated node which represents the shared VP components and local components for that particular rendering device. In Figure 3A, VP1 VP21 VP3 and VP, represent the versions of the virtual proof for each rendering device. The arrows from SOM 13 or imagical 14 represents the measurement data received by the node incorporated into color calibration data, which is stored in a local component of the VP.

The VP provides system 100 with many useful features, which include remote proofing for both intermediate and final approval of color products, conferencing at multiple nodes in the network between users which may have different rendering devices, and distributing color preference data with or without page layout image data. The conferencing mentioned above allows users to negotiate over the colors appearing in page layout and to confer about color corrections. For example. conferencing may use video displays 17 (soft proofs) of the page layouts using remote annotation software, such as imagexpo. Another important feature of the VP is that it is modifiable such that as changes occur at a rendering M.

device, such as in inks or substrates, system 100 can automatically adjust the rendering device's cz - -)ration. In addition. adjustments to calibration may be performed on demand by a user. This allows a user to update color preferences, such as color assignments of page layouts being rendered by rendering devices in the network without retransmitting the entiret of the imaize data.

A feature of system 100 is that it compensates for differences in the gamuts of different devices. As described earlier, a gamut is the subset of humanly perceivable colors that may be captured or rendered by a device. The preceding definition implies that the ideal or limiting gamut is the set of all colors that a normal human can see. It is important to distinguish between receptive and rendering gamuts. The former refers to the gamut of a sensor or camera of a CNII, or human. The latter refers to the colors that an output rendering device is capable of producing in a medium by application of its colorants. Although it may be possible to design a rendering device that may produce all the colors we can see under suitable circumstances, rendering gamuts are generally smaller than the human perceptual gamut due to the properties and limitations of practical reproduction media. For example, the gamut of a color print viewed in reflected light is generally smaller than that of a video display device viewed in controlled illumination which is generally smaller than the gamut available with positive photographic transparencies. All the foregoing rendering gamuts are generally smaller than receptive gamuts.

The CNH's of Figures 3B and 3C are colorimeters such as discrete (unitary) colorimeters (SOM 13) or imaging colorimeters (imagical 14) which may be used in conjunction with single-channel light-detectors. These colorimeters may be stand-alone units or built-in to a rendering device. As stated earlier. the CMIs are controlled by their associated nodes in system 100 for calibration and verification of rendering devices. SOMs 13 and imagicals 14 are specialized to measure color in different ways. SOMs are suited for measuring a spatially homogeneous patch of color, preferably in a multiplicity of spectral bands. Preferably, at least 15 spectral bands spanning the visible spectrum are sampled. making a SOM more-than-3-channel input device. The transformation from relative spectral energy or reflectance to image colors employs the convolution, a similar technique is described in CIE Publication 15.2, page 23. cited earlier. An example of a SOM is a unitary colorimeter or a spectrophotometer of a U.S. Patent No. 5,319.4337 issued to Van Aken et al.

However, the distinction between SOMs and imagicals is not intrinsic. A SOM with a sufficiently restricted aperture and area of view and which could perform the spectral integrations sufficiently rapidly and which could scan rasters of image data may qualiy as an imagical. Imaizicals are suited for multi-scale (i.e.. variable resolution) imaging colorimetry.

1 1 consisting of an array of photosensors, such as CCDs, capable of sensing color, as would the Standard Observer.

SOM 13 is calibrated against a reference standard illumination source whose calibration is traceable to the U. S. National Institute of Standards and Technology or similar organization. The calibration of SOM 13 is generally set by the factory producing the device. SOM 13 should be periodically. recalibrated to assure its reliability. Calibration of imagicals will be described later.

Further, SOM 13 may be used in conjunction with an imagical. SOM 13 can provide a check on the calibration of imagical 14 by sampling some of the same colors measured by the imagical and providing reference data to compare against the imagical measurements. Under suitable circumstances the SOM enables a spectral interpretation of what is seen by the imagical so that a spectral illumination function characteristic of viewing can be substituted for that used in measurement as described in corinection with Figure 3D.

Referring to Figures 3B and 3C, preferred configurations for sensors for CMIs in system 100 are shown. Because colorimetric accuracy of CMls and their ease-of-use are desireable, the preferred configuration of CMI device colorimetric sensors is to have them be as unobtrusive as possible to the user. Thus, in the preferred embodiment, CMIs are built-in to the rendering device.

In the case of a conventional video display or monitor, Figure 3B shows the preferred embodiment for sensor of a CMI. A cowel 26 attaches to a upper chassis 27(a) to frame a faceplate or screen 24 of video display 22 and to shield the display from most ambient illumination. A fiber optic pickup (not shown) coupled to a projection type lens or lens system 28 plugs into cowel 26 in order tomea-sure color of screen 24 without actually touching the screen or requiring placement by the user. Path 30 shows that the line of sight of the lens and fiber optic pickup reflects off faceplate 24 and views the blackened inner surface of a lower cowel 32 attached to lower chassis 27(b) such that it does not see specularly reflected light reflected from faceplate 24. Nonetheless. operation of display 22 is preferably in an environment with subdued illumination.

Preferably the CMI for a display screen is as a unitary colorimeter SOM 13. The unitary colorimeter takes color measurements via lens system 28 when needed in response to instructions from circuitry at a node. Unitary colorimeter SOM 13 can measure relatively small areas of screen 24 using a sensor connected to the fiber optic pickup. This sensor can be a spectral sensor, a 3 or 4 filter colorimeter or a single channel sensor. A spectral sensor must be able to resolve 2 nanometer wavebands across at least the long wave (red) end of the spectrum in order to make an adequate measurement of the red phosphor of conventional CRTs. This could be accomplished with a grating monochromator and linear photodiode array, or linearly variable interference filter and diode array. Scanning the red end of the spectrum may be performed with a narrowly, electronically-tuned, spectrally variable optical filter in order to locate red phosphor peaks exactly. If a 3 or 4 channel filter colorimeter is used, compatibility with system 100 requires that the spectral sensitivity of the arrangement be a linear combination of the human color matching functions with acceptable precision. The video display 22 phosphors change on a slow time scale. Therefore, provided that the primary chromaticities of the display are measured on a regular schedule, the SOM of Figure 3B may be replaced by a single-channel meter for routine measurements of gamma (the voltage in - photons out characteristic of the device) very accurately for the three channels.

Alternatively, an imagical may be used instead of a unitary colorimeter SOM 13 in Figure 3B. In this case the sensor of the imagical may be centered in a door (not shown) which is attached to cover the aperture 29 of cowel 26 and 32 (the cowel may surround the entire perimeter of the chassis) so that the sensor views faceplate 24 center along a line of sight. Imagical may acquire data needed to flatten the screen, i.e., make it regionally homogeneous in light output or to compensate for interactions among adjacent pixels in complex imagery. However, both of these factors are second-order effects.

It was noted earlier that the preferred embodiment utilizes a video display 17 that projects the image, preferably onto printing paper. In this configuration, the CMI which monitors output is situated near thesource of projected light. as close as possible to being or. a line of sL, . Preferably, a projection video display is capable of resolutions exceeding 1200 lines per inch. In this way. it is possible to simulate rosettes. moires and details of image structure on the hard copy and to show the actual structure of images captured frorn prints with a CMI equipped with macro optics. To provide high resolution projection. butting together a composite image using several I imi ted-area-disp lays may be performed. Regardless 1 of type of display 17, the display may use additional primaries in order to extend 9=ut or better simulate subtractive colorants.

Figure 3C shows an example of a sensor of a CMI for measuring a substrate 34 (hard proof), such as produced by a proofer device. In the case of a digital proofing device, such as a dye sublimation or ink jet printer, it is desirable to position a dual fiber optic pickup/illuminator 36 arrayed for 45 and 90 degrees measurement geometry near the print head in order to sample recently printed color pixels of substrate 34. Pickup/illuminator 36 have projection-type lenses coupled to both a sensor fiber and a illumination fiber within a light shielding sleeve to implement non-contact measurements and to protect against stray light. Pickup 36 relays light to an analysis module of a CMI in which, preferably. a spectral colorimetric measurement is made. If placement near the print head is not practical. then placement over the exit path of printed sheets is preferred. This sort of placement is preferred for proofing devices which are page printers, such as electrophotographic or conventional. analog proofers. As was the case with monitors, the use of a single-channel device to monitor linearization is compatible, provided that the nature of the device variation over time is compatible with intermittent, full, colorimetric calibration. For example, it may be sufficient -to calibrate a dye sublimation proofer only once for each change of ribbon. Although it Is preferred to build the CM into the proofing device in order to minimize user involvement with the process, system 100 can alternatively require the user to place the printed copy on a X-Y stage equipped with the dual fiber optic pickup/illuminator 36.

At production node 104 of system 100, it is preferred that images rendered from a press are captured as soon as possible after they are printed, i.e., after all the colorants are applied. In this way the data necessary for control are available promptly. Although system 100 may estimate the color-error of printed sheets relative to an aim value. the preferred embodiment applies imaging colorimetry of a CMI to the analysis of the image area. Preferably. there is a spectral component to the CMI measurement such that interactions of the colorants with the stock or printing substrate can be analyzed and it Is easier to compute the colors as they will appear under standardized viewing conditions.

Imagical 14 of a production node 104 may employ SpectraCube technology. reference cited earlier. or cameras using any filtering technology that can simulate Standard Observer response adequately. The preferred embodiment, imagical 14 has one or more cameras, such as solid state area arrays, in which the light can be filtered through at least three wavebands, either simultaneously or sequentially, in conjunction with a unitary type colorimeter which views a determinable region of the image viewed by the camera. This provides for the possibility of a spectral interpretation of each pixel as needed. A spectral interpretation is desirable in order to be able to control color to a criterion of how it will appear under the final viewing illuminant. For example, an on-press camera could be several cameras/sensors. each equipped with a relatively narrow-band filter. The cameras, used in conjunction with a unitary colorimeter. can sample composite spectral curves, as shown in Figure 3D, in such a way that inference of the total spectral reflectance function is possible. The minimum number of camera channels depends on the number of colorants and the complexity of their spectral absorption curves. The cameras may include an infrared-sensitive camera to differentiate the black contribution to a spectral curve from contributions of non-neutral colorants.

Preferably. imagical 14 is capable of multifocus or variable focus imaging so that larger areas of the image can be viewed at lower resolution or vice versa. Views of small areas at high resolution enable simulation of fine image structure by proofing devices capable of sufficiently high resolution display. It is also preferred that imagical 14 is equipped with anti-aliasing filters since much of the input to the imagical can be expected to be screened or pixellated. The article by Grievenkamp (cited earlier) describes an example of anti-aliasing. Also preferably, the viewing illuminant is controlled to simulate the viewing illuminant during the measurement. This may be achieved using a spectral 1 y-adaptive, electronically tunable filter to match the illumination spectrum of any desired light source. which is described in references by Shnitser and Hoyt (cited earlier).

Ease of use and accuracy requirements indicate that CM1s are calibrated or self calibrating in the preferred embodiment. The preferred embodiment of the unitary device SOM 13 approximates a dual-beam device. such as spectrophotometer of Van Aken et al. cited earlier. The spectrum of light reflected from the unknown sample is compared either simultaneously or successively with the light of the same source reflected from a known reflector. In this way it is possible to separate the spectral contributions of colorants. substrates and illumination sources and to estimate their true functional forms over multiple 1 impressions. This increases the CNU's accuracy. Even with such measuring, however, regular recycling of instruments for factory re-calibmtion or field recalibration using standardized reflectance forms should be performed. Also, if video display 17 is a selfemissive monitor, the above dual-beam functionality although not useful in computing a difference spectrum, provides a means of evaluating possible drift or need for re- calibration in the instrument.

System 100 operates in accordance with software operating at the nodes, which is preferably based on object oriented coding, a well known programming technique. However. other programming techniques may also be used. The discussion below considers the different input and output devices, e.g., CMIs and rendering devices, as objects. Each object refers to software applications or routines in system 100 which provides input to or accepts output from other applications (or devices).

Referring to Figure 4A, a three-dimensional matrix model of the abstract class device/profile is shown. For example, a device may be instantiated as a linear input device or non-linear output device and derive properties through inheritance. An object created from a certain class is called an instance, and instances inherit the attn'bute-s of their class. Inheritance is a functionality of object-oriented coding and provides for grouping objects having common characteristics into a class orsubclass. The matrix is extended in a third dimension to include ')-colorantchannel- devices -and more-than-3-colorant-channel devices. Display devices may have 4 or more colorant channels (for example, Red, Green, Cyan. Blue. indicated by RGCB, Monitor 38) in order to enhance gamut or better simulate subtractive color reproduction processes. In other circumstances, a display device may appear to fall into the same class as a CMYK printer 39, which it represents in a soft-proofing/client relationship. The appropriate models and transforms are those associated by inheritance with the subclass into which the new device falls within the general matrix.

A CMYK printer 39 exemplifies the (sub)class output/non-linear/more-thanthreechannel. By inheritance. client-proofer relationships are shared by members of subclass output. since the ability to enter into clientproofer relationships can be passed to all members of the subclass by inheritance. Note that the subclass of input devices is distinguished b%. conservation of gamut among its linear members. Likewise, specific instances of subclass 28 - linear inhcri t an association with linear matrix models of color transformation, whereas nonlinear subclass members associate with color transformation by polynomial evaluation, interpolation or other form of non-linear color mixture function. Procedures performing the color transformations are incorporated in the data structures defining the objects. and are discussed later in more detail. More-than-three-channel devices require procedures for rendering with the extra colorants, which is also discussed later in more detail.

Note that the class hierarchy depicted herein supports instantiation of devices of types "scrir" I'mritr" and "prtr" as promulgated by the prior art ICC Profile Specification (cited earlier). However, the class hierarchy disclosed herein is considerably more general, flexible and extensible. The properties of flexibility and extensibility are illustrated by the following practical example: a video display ("mntr" in prior art ICC Profile Spec) in system 100 may occupy any of a number of cells within the class strucrure depending on its physical properties (for example. the number of colorant channels it has) and purpose (stand- alone design workstation or soft proofer representing a four-colorant device.)

The term "linear" herein, when applied to a device, means that linear models of color mixture can successfully be applied to the device (see previously cited art by Gordon and Holub and by Holub.) It does not imply that a video display is intrinsically linear. For example, among input devices, linear defines that linear color mixture models can be used to convert from CIE TriStimulus Values to linearized (or gamma compensated) device signals and vice versa. Further note, that one application can be an output device with respect to another. For instance, application software may convert RGB TriStimulus Values into CMYK colorants and occupy the same cell as a CMYK printer 39.

The calibration of devices in system 100 is indicated by the classes of the devices 40. 41, 42 and 39 in Figure 4A. Calibration herein includes the process of obtaining color transformation data in uniform color space. Once devices at nodes in a configured network 11 of system 100 are calibrated, the colorants produced by nodal rendering devices can then be controlled. however such calibration remains subject to recalibration or verification processes, as described later. There are four classes of devices 40. 41, 42 and 39 which require calibration:

I) imaging colorimeters or imagicals 14 (class input/11 near/3 -channel); 1 29- 2) video displays 17 (generally in class output/linear/3-channel), 3) unitary, spectral coiorimeters or SOM 13 (class inpuvlinear/more-than3-channel); and 4) printers or presses (generally in class output/non-linear/more-than-3 - channel).

Optionally, non-linear input devices may be used in system 100, as described in the article by Sodergard cited earlier, but are less preferred. An example of a procedure for calibrating non-linear input devices to a colorimetric standard is described in Appendix B of the American National Standard "Graphic technology - Color reflection target for input scanner calibration" (ANSI IT8.7/2-1993).

In the first class of devices, the calibration of imagicals 14 involves preparing compensation functions for the separable non-linearities of the device's transfer function. which are usually addressed in each color channel individually. These compensation functions may be realized in onedimensional look-up-tables (LUT), one for each color channel. The compensation functions may be defined in a calibration step in which measurement signals from imagical 14 are generated in response to observation of step wedges of known densities. Next, specifying the constant coefficients of a linear color mixture model expressed as a 3X3 matrix transformation. hereinafter referred to as matrix M. which converts linearized device codes into CIE TriStimulus Values, such as XYZ, or related quantities. The formation of matrix M is described in Gordon and Holub (cited earlier). Last, the gamut of the input is scaled to fit within the color coordinate scheme in which images are represented. Because inputs to the system are printed copy (proofs and press sheets,) gamut scaling is often unnecessary except when the image representation is a space such as calibrated. monitor RGB which may not encompass all the colors of the print. Occasions on which this would most likely be a problem are those involving extra coiorants used for "Hi Fi" effects. although limited ranges of conventional printing cyans and yellows are out-of-gamut for some monitors. Preferably. imagicals 14 are selfcalibrating to assure the accuracy of their color measuremenES. The compensation function LUTs. matrix M. and possible gamut scaling data are considered the calibration transforms for each imagical 14.

The color transformation in imagicals 14 into uniform color space based on the above calibration is generally shown in Figure Q. Imagicals 14 output measurement signals R. G and B, also referred to as device codes. The measurement signals are passed through compensation function LUTs 48 (or interpolation tables) to provide linearized signals G,...

and Bli, in order to!.nmpensate for any non-linearities in the relationship between light intensity sensed by imagical 14 and the device codes. Matrix M then operates on the linearized signals Gli,, and Ri,, to provide X, Y, and Z coordinates. Matrix M is shown consisting of the 3x3 coefficients (a00 -,) defining the linear combinations of k, Q, and B,,,, needed to match X, Y and Z, the TriStimulus Values (TSVs) of the CIE Standard Observer. Although not shown in Figure 413, gamut scaling is performed after TSVs are converted to Uniform Color Space coordinates such as those of CIELAB. Scaling of an input gamut onto an output gamut in this case is entirely equivalent to processes detailed for rendering devices later in this description.

Calibration of video displays 17 in the second class of devices follows the same steps for imagicals 14 described above, however since video displays 17 are output rendering 0 devices, matrix M and compensation function LUTs are inverted. Calibration of video displays for soft proofing is well known, and discussed by Holub, et al. (J. Imag. Technol., already cited.) Referring to Figure 4C, -k device independent coJor coordinates XYZ are input signals to display 17. Rendering employs the inverse of the operations used for imagical 14. The inverse of the calibration matrix M is called A (to emphasize that we are considenng numerically different matrices for the two devices) and is used to convert the XYZ input signals to linear device signals Wii, G'Ii,, and W,,, The linear device signals R,,. Q,., and B',i,, are postconditioned using the inverse of the compensation function LUTs which define the non-linear relationship between applied signal and luminous output of display 17, a function which is defined and adjusted in a separate, empirical step of calibration. The output "I p J -t t-o from the LUTs are gamma corrected signals R". GY. and BY. representing the'iuFainees cc display 17. Note that there is no necessary relationship between the matrices A and M in Figures 413 and 4C. Further, since the LUTs of Figures 413 and 4C may be used with various types of transformations in system 100. they are preferably represented by a separate data structure. within the software architecture. which may be combined like building 1 31 - blocks with other structures. such as 3X3 matrix, or multidimensional interpolation table, to form more complex data structures.

As stated ca.-'..':-r video displays 17 generally belong to the subclass output/linear/3channel in Figure 4A. However, there are two important exceptions to this:

a) when display 17 is used to represent a more-than-3-channel printer, the transforms used to drive the display are non-linear -- in other words, a proofing device manifests attributes of its client; and b) when display 17 is used as an accomplice in the creation of computer-generated art, the video display can be considered a linear input device, since new. digital, RGB data are created in the medium and color space of the display. Likewise, calibration for a RGCB Monitor (device 38 of class linear/output/>3-channel) is a simplification of procedures for calibrating the class of non-linear/output/> 3-channel devices. which is described below.

In the third class of devices, the calibration of unitary, spectral colorimeters or SOM 13 is set by the factory. as discussed earlier, and thus does not require the preparation of calibration data to provide device independent color coordinates.

Referring to Figure 5, the process of calibrating the"fourth -class of devices is shown. It should first be noted that for proofing devices in order to represent one device on another (to proof) it is necessary to have an accurate model of the color reproduction of both devices. It is preferred that the proofing device has a larger gamut than the client and that accurate knowledge of gamut boundaries be derivable from the models of colorant mixture. Because proofing is an issue with output devices, it is necessary to develop rendering transformations that invert the models of colorant mixture and that mix colorants on the proofer in such a way as to match the colors produced by the client as closely as possible. In other words. this should involve respecting the gamut limitations of the client device when rendering on the proofer. In summary. the following four kinds of color transformations are developed in calibration of the fourth class of devices:

1) forward models that enable calculation of the color. in device independent coordinates. of a mixture of colorants; 2) forward model inverses that enable calculation of the amounts of colorants needed to render a desired device independent color coordinate-, 3) descriptions of gamuts in terms of boundaries specified in device independent color coordinates; and

4) mappings of colors realizable on one device onto those realizable on another in a common, device independent coordinate system (g=ut configuration data). The above four color transformations are discussed in more detail below. The following is in reference to a hard copy proofing device or proofer, but is applicable to other devices in the fourth class. including high and low volume presses.

Step I of Figure 5 is the process of preparing linearization functions, which is shown in more detail in Figure 6A. This process establishes a linear relationship between the digital codes sent to the proofer and the output of the proofer. measured in units such as visual print density. Linearization improves the utilization of the available digital resolution and is usually implemented by means of one-dimensional Look Up Tables (LUTs) that map linear digital codes from the nodal computer onto signals that drive the marking engine of the proofer to produce an output that is approximately linear. For example, step wedges printed in C, M. Y and K respectively should produce gradations in measurable visual density that increase linearly as a function of the digital codes commanded by the host.

Usually, linearization involves printing a step wedge on the marking engine without the benefit of the LUT - if data are passed through the LUT, it applies an identity mapping. The color samples of the wedge are analyzed by the CMI associated with the proofer and the measurements are supplied to the nodal processor so that it can compute the transfer function from command codes to print densities. The measured transfer function is compared to the desired one and a function that compensates for the errors in the measured function is prepared - this is what is loaded into the LUT for use in normal image transmission. The LUTs are written to the local part of the VP as linearization functions.

Linearization is not a strict prerequisite for the remaining procedures because a multidimensional color transformation could be desiened to accomplish what the onedimensional LUTs are supposed to do. Thus. linearization of step I although preferred in system 100. may optionally be incorporated into other color transformations in Figure 5. However. it is generally advantageous to exclude as many sources of non-linearity as possible from the transformation function which is specified by the procedures outlined here.

1 33 - Step 2 of Figure 5 involves verifying or renewing the calibration of the CNO and rendering calibration forms, and is described in the flow chart of Figure 6B. After initializing calibration procedures, if the CNE associated with the rendering device is an imagical 14, it is calibrated to provide calibration transforms, as described above. Preferably calibration of the CNfl is performed automatically in response to instructions from circuitry at the node.

After the CMI associated with the rendering device is calibrated, a calibration form is rendered on the proofer. For example, this form may have the following attributes: I.) a sampling of all combinations of four levels of all of the colorants, 2) inclusion of approximately neutral step wedges, 3) several samples of flesh tones, 4) a number of redundant patches - these are common inkings placed at different locations on the proof to provide information about spatial non-uniformities of the proofing process. It also is useful to include at least short step wedges in each of the colorants and their overlaps. for example. cyan, magenta and yellow as well as blue (cyan + magentaJ green (cyan + yellow) and red (magenta + yellow.) The calibration form described consists of about 300 samples in the case of 4 colorants and can fit on an 8.5 X I I inch (21.5 X 28 cm) sheet with patch sizes of I cm. on a side. However, generally, the number of patches should be three times the number of polynomial terms fitted to the data (discussed later in step 4. Figure 8.) Patch sizes are scaleable for compatibility with various CMIs. In addition to the tint samples, the target has markings similar to pin registration marks and demarcated handling zones to facilitate transfer of the target from the proofer to the CM if the CMI is not incorporated into the proofer. The registration marks indicate clearly where and how to insert the copy in a stand-alone CMI so that the instrument will find the patches where they should be and the handling zones will emphasize to the user that the image area should not be touched. Hard copy proofers may write an identifying number and/or bar code for the device and for the specific proof Nncludin -ate and time) on the proof.

Alternatively, a device in the fourth class. such as a running press. may be calibrated bv analysis of live imagery (rather than a calibration form) digitized by an imagical provided 1) that the images analyzed sample the gamut of the device adequately and 2) that the effects of page adjacency Within a signature on color reproduction can be accounted for (for exampile 4 by reference to stored historical data at the node.) As always, the relevant data for calibration are the known colorant specifications embodied on the printing plate and the colors resulting on the printed sheets.

After rendering the calibration form, the form is measured by a CMI and calibration data is provided (Step 3 of Figure 5). The processes of step 3 are detailed in the flow chart of Figure 7. As stated earlier, the preferred CM is a linear colorimeter, such as a imaging or unitary colorimeter, which is capable of providing spectral analysis data which supports illuminant substitution. The CMI should produce several readings of color to the node. if one is an outlyer compared to the others because of a blemish on the copy or some malfunction, then the software at the node excludes it from the averaging for that patch. If no two of the measurements agree. then the patch should be flagged so as to identify it as a problem in subsequent processing. More generally, the number of measurements of each patch is odd to permit voting by the software in the interest of selecting trustworthy measurements.

If imaging of the calibration form is performed by an imaging colorimeter or imagical 14, then the imaging colorimeter analyses the images on the form, and uses its calibration transforms from step 2 to calculate image colors and standard error of measurement inCIE Uniform Coordinates, such as L'. i. b. Also the imaging colorimeter provides many sample values of each patch color; regional checks for nonuniformity of color in the sampled area should be performed. However, if imaging of the calibration form is per-formed by an unitary colorimeter or SOM 13, then the patch readings from the form are converted to color measurements in CIE Uniform Coordinates.

Each patch measurement may include information about sensitivity, integration time. wavelengths sampled, illuminant substitutions, and the like. The series of measurements from each calibration form are accompanied by at least one record of the reference spectrum. although. obviously, reference spectral data will be collected and used on every reading. at least for reflection measurements.

Regardless of the type of CMI, a list of colorant values (the Independent Variables. IVs. to the fining procedure) to corresponding color values (Dependent Variable) with standard deviation are assembled. On this list of measurements outlyers are flagged. An estimate of sheet variation from redundant sampling is produced.

1 - in addition to multiple measurements within a given sheet, two other means for enhancing the reliability of the data are provided. First, the software supports measurements of multiple sheets and second, an historical record of measurements firom a particular proofer or press is preferably maintained at a node. Historical data can be stored more compactly and compared more readily to current data if the measurements are converted from spectral form to colorimetric form. Spectral data of the measurement is stored in a database at the node in terms of color and summary statistics.

Preferabl, the database maintains a FIFO history of color and summary statistics from r f: c ent. v the most'feeent measure forms. Because step 5 involves least squares error minimization, flagging of outlyers is preferred to reduce the influence of one bad reading. A decision on whether a current reading is legitimate is made by comparing CIE AE values rather than the two spectra. The assembled list with flagged outlyers and standard deviation of each patch measurement is written to a calibration (cal.) data file in the local part of the VP, for later use in building the forward model.

After step 3 is complete, processing continues to step 4 of Figure 5, building a forward model based on the calibration of step 3. Step 4'is flow charted in-Figure 8.. This model will represent color as a function of colorants of the proofer or press. It should first be noted that generic polynomials provide a satisfactory form for models of color generation by colorant mixture on printing devices. However, this does not exclude any other mathematical or physical modelling procedure capable of producing transformation functions of sufficient colorimetric accuracy. Polynomials of relatively low order in each of the colorant variables may be fitted very nearly to within the device variation. In other words, the uncertainty of the model prediction is not much greater than the uncertainty of color rendered in response to a given set of digital codes. Low order implies that the colorant variables are not represented as powers greater than 2 or 31 and that the sum- of the powers of the independent variables in a single term of the polynomial is limited to 4. - Therefore. if the inks C, for cyan. M. for magenta. Y. for yellow. and K. for black are the independent variables. a valid term could be C2MK.but not CW K. Analytical derivatives are easily computed. which makes them advantaecous for model inversion.

A polynomial forward model is fitted to a data set consisting of independent variables of colorant and dependent variables of device independent color coordinates (stored in the calibration data file of the VP) by the method of least squares. A polynomial is a linear combination of each of its terms which are called. mathematically, basis functions at step 82. Without loss of generality, discussion can be simplified by considering functions of two variables, C and M, in which each variable may be raised to a power of up to 2 and the sum of powers may not exceed 4:

+ GOIM t 01,CM + G"CIM + 00_ - 012CM2 +,O""C2M2 R= aOO + a j()C + aioV jM2 G= boo bjOC + b 3OC2 - bOjM + bjjCM + b2jC2M + b02M2 + b 12CM2 + b,,C2M2 B= c,, c,,C + C,oC2 + CojM + C11CM + C,,C2M + co2W + C12CM2 - c, C2M2 in the foregoing. color is the vector valued function ', R G B 1, of the variables C and M and the a's, b's and c's are constant coefficients which give the proportions of their corresponding terms to be mixed in fonnine the linear combination. The purpose of the fining procedure is to find the set of coefficients which results in the least squared error when comparing the color measured at a given patch with the color calculated by substituting the patch's colorant values into the forward model. The r)olvnomials are untruricated within the constraints on the powers. The DV may also be L, a, b coordinates of CIELAB uniform color space.

In Figure 8, the process of building a forward model begins by building a design matrix for the problem (step 82). The design matrix has M columns, one for each basis function, and N rows, on-- for each patch measurement. The n=be,- of rows should exceed the number of columns; practically, the ratio of N to M should be areater than 3 for szood results. After reading the cal. data file, each cell in the matrix is filled with the value of tht basis function fo r the column at independent variables (inks) given by the row, divided by the standard deviation of the patch measurements, if available. else by I (step 83). Note that the patch measurements themselves do not enter the design matrix. Then use the design matrix and vecror- patch measurements for each of the three, color, dependent variable dimensions to write matrix equations that can be solved for the desired coefficient vectors preferably by Singular Value Decomposition. SVD (step 84).

The numerical methods outlined in the preceding and the following paragraphs are similar to those described by Press, et al. urnerical Recioes. Cambridge University Press.

1 Cambridge, UK, 1986.) Sections 14.3, "General Linear Least Squares," with SVD fitting, and 0 5.3, "Polynomials and Rational Functions".

The model and its derivatives can be evaluated efficiently by a method of recursive factorization. The method depends on use of polynomial terms as basis functions. However, it permits evaluation of the function with no more than I multiplication and I addition per polynomial term. Because the independent variables never need to be raised to a power, the demands for precision on the computational machinery are not as great. The principle can be seen most readily in one dimension; the function y = k + a,x + a-r' can be factored to yield a + x(a, + ar) which evaluates with 2 multiplies and 2 adds. Generalizing this to the twodimensional flinction. described above: a0c + t710C + az(,C7 - M(aal + a,,C + a21C) + M'(a., - a12C + o,,C'), or a00 + C(ajo + a20C) - M[(aol + C(a,, + a21C)) + (aQ2 + C(ct2 + a,,C))Nfl. How to eeneralize to three or four dimensions is apparent.

At step 8 1, the patches that were flagged during measurement as outivers are excluded from the fitting. The fitting program estimates the variatign in thg device based on deviadon$. of color in redundant patches and/or measurements of multiple copies. The average -error of fit is calculated as the average AE' (CIE Color Difference Unit) over the set of measurements of patch colors compared to the colors predicted by the fitted polynomial (step 83). This average should not exceed the estimated device variation by rnor-- than I nE unit. %Vhen i: does, or when the software detects individual patch discrepancies exceeding 4 or 5 &E units, the software flags apparently outlying points (step 85). It then computes trial fittings with - - - 6--, outlyers omitted to see if average error can e improved (step 86). It also is informed with a strategy for applying various sets of basis functions in the interest of achieving an acceptable 0 - fit. However, the firting procedure will reject a dataset at step 86 rather than Let too involved with it.

Principal Component Analysis, or an equivalent method. may be employed to reduce the number of polynomial terms (complexity of the model) consistent with -a giVen avera2e error criterion. this technique is similar to those described in Johnson and %VLChL-rn. -k[)[)i1:d NlulcivariaEt Statistical AnalvsLs. 3rd ed., Englewood Cliffs, NJ: Prentice Hall. 1992, ch. S.

38 - Fitting concludes with writing a final polynomial model descriptor (a data structure and a file) consisting of a header and lists of coefficients. The header includes information such as the identity(ies) of proofs measured, relevant dates and times, the ingredient data files, the form of the polynomial (maximum powers of the independent variables and maximum order of a term) goodness of fit statistics. The polynomial descriptor will be needed later by the polynomial evaluator of the software and is written into the shared portion of the Virtual Proof.

Although this discussion is directed to a 4-colorant printer or press. the polynomial forward model is, potentially, part of a soft proofing transform which enables a video display to represent a client printer. It can be used to compute a colorantA to colorantB transformation for what is effectively a CMYK. subtractive-color video display. in that CWK colorant values are processed through a transformation to produce device- specific RGB for a monitor. This can generalize to use of more -than-fourcolorant-printers and more-than-three-colorant displays.

After the polynomial model descriptors are computed, a forward model table (FMT) and prototype gamut descriptor data are prepared (step 5 of Figure 5). The processes of step 5 are detailed in the flow chart shown in Figure 9A. As described above, the polynomial based on the model descriptors represents the forward model. The forward model enables us to predict the colors resulting when certain colorant mixtures are asked for on the printer or press. System 100 includes a polynomial evaluator to evaluate this polynomial either in hardware circuitry or software at the node. Referring to Figure 9B, a topology of operators is shown (multipliers 86 and adders 87) to implement the polynomial evaluator for the twocolorant case, cyan (c) and magenta (m). Each operator receives two inputs. Inputs (operands) designated by two numerals are the constant coefficients and inputs designated by C 89 or M 90 stand for the independent variables. amounts of colorant. For instance. 21 88 denotes the coefficient which corresponds to the second power of C times M. In order to evaluate a function of three variables. 3 of the units shown in Figure 9B are needed. Four variables require 27 such units for an untruncated polynomial. In a hardware realization of the evaluator, it is preferable to retain the generality of the polynomial form and zero poly, terms (or their coefficients) that are missing due to truncation. If 27 units are too many for a 1 39- cost-effective device, then the calculation may be staged and the successive stages pipelined through a subset of the 27 units, albeit at some cost in control logic and speed. Given the opportunities for parallelism in a hardware implementation. the preferred embodiment benefits by a chip to transform ink values to colors at video rates. Such a chip may be a component of nodal circuitry encompassing a graphics accelerator to drive a video display device for soft proofing at the node. It accelerates the generation of color separation transformations because evaluation of the colorant to color model is a major factor in that computation. It also accelerates the evaluation of color separation transformations in situations where the data of the inverse, color-to- colorant conversion can be fitted by polynomials with a sufficiently small average error of fit. Data for fitting can be the addresses and entries of an interpolation table of sufficient size, as discussed below.

The FMT stores the results of the polynomial evaluator in a data structure and colorant quantization/addrcssing scheme shown in the CMYK hypercube of Figure 9C, which contains the entire gamut of colors reproducible by a CMYK printer. The hypercube may be subdivided into a FMT having 17 points (16 intervals) per colorant dimension for sufficient accuracy. However, the software architecture supports more or less, within the constraint that the numbers of grid points per dimension satisf the equation 2'+1, where n is integer. Depending on the requirements of the rendering applications and equipment, a software switch controls whether the tables are written in pixel interleaved (each address of a single table holds the values of all dependent variables) or frame interleaved format. In the latter case, three M-dimensional tables are prepared, where M is the number of colorants. Each "cell" of a table has an M-dimensional address and contains a color coordinate computed by evaluation of the forward model at step 97 of Figure 9A. At step 97 nested looping over all colorant addresses is performed and computed colors from the forward model are stored at each address. Thus, each model evaluation yields three color coordinates each of which is deposited in a cell corresponding to the M colorants which produce it in the appropriate table. Colors of inkings in the midst of a hypercuboid are estimated by interpolation.

Referring now to Figure 9D. each of the M channels of input to the FMT may be provided with preconditioning LUT for each of the Independent Variables and each output channel may be processed through a 1 -dimensional postconditioning LUT. interpolation is preferably performed by linear interpolation in the nodal circuitry or software.

The data structure in Figure 9D accommodates the preconditioning of j address variables byj 1-dimensional transformations which may be implemented as look-up tables with or without interpolation 98. The preconditioning transformation may pass one or more of the j inputs Independent Variables (TVs) through to the multidimensional transform unaltered (identity transformation) or may apply a functional mapping such as a logarithmic conversion. The multidimensional transform 94 has j input variables and i outputs. The preferred implementation of the transformation is by interpolation in a sparse table of function values or by evaluation, in hardware, of a set of polynomials fitted to the tabular values (where firting can be done with sufficient accuracy.) In Figure 9D. a 3-dimensional TV 93 is applied to the multidimensional transform 94. Multidimensional transform 94 consists of many smaller cuboids 95 whose comer points are the sparsely sampled values of the IV at which values of one (or more) of the dimensions of the dependent variable, DV, 96 are stored. The IV provides addresses and the DV contents. The subcuboid 95 is shown at the origin of the addressing scheme. Interpolation is used to estimate values of the DV occurring v at'volue of the IV which are between comer points.

The data structure of Figure 9D also accommodates the post-conditioning of the i output variables (DVs) of the transformation by i 1-dimensional transformations which may be implemented as look-up tables with or without interpolation 98. One of the transforming functions which may be included in the post-conditioning is the linearization function optionally produced by step I of Figure 5.

The purposes of the]-dimensional pre- and post-conditioning functions include improving the extent to which the relationship between variables input to and output from the muffidimensional transformation 94 is approximated by whatever interpolation function is employed in evaluating the transformation. Therefore. the form of the functions should be known from preliminary studies of the device and the pre- and post-conditioning transforms must be in place during calculations of Steps 2 through 9 if they are to be used at all. For example. a linearization function which may be defined in step I should be used in rendering the calibration target in step.1.

1 A linear interpolation formula for two dimensions which generalizes to higher dimensions with a simple, proportional scaling in the number of operations is described in Gallagher, "Finite Element Analysis", cited earlier. For example, given a cell in a twodimensional array of sparsely sampled points as shown in Figure 9E, the interpolated value of a function f(x,y) at a point (x,y) interior to the cell may be calculated as the weighted average of the endpoints in the direction of the longer of the components x or y plus the weighted average of the endpoints in the direction of the second longest component. In other words. order the distances of the point with respect to the axes of the cell and then sum the fractional distances along those ordered dimensions. Written as an equation: for v < x, f(x,y) f(0,0) + x (f(1,0) - f(0,0)) + y (f(l, I) f(1,0)), and for x ≥ y, Rx,y) f(0,0) + y (f(0,1) - f(0,0)) + x ff(l,l) f(0,1)).

Figure 9A also flowcharts the preparation of the prototype gamut descriptor (GD) data. As colorant addresses are converted into colors to populate the FMT (step 97). the colors are also converted into cylindrical coordinates of CIE hue angle. chroma and lightness. Hue angle and lightness are quantized to become addresses into a 2-D gamut descriptor. preferably dimensioned at least 128XI28 for adequate resolution (step 99). The Chroma component of the color is not quantized, but is stored in linked lists of Chroma values associated with each hue angle, lightness coordinate. The finished gamut descriptor data is prepared from the prototype GD data later at step 7 of Figure 5, and consists only of surface chroma values for each coordinate. Therefore, the prototype is not a shared file and is written to the local portion of the VP, while the FMT is written to Elie shareable portion.

Next, at step 6 of Figure 5, the FMT is inverted into a prototype transformation table. also called Proto SEP table. Rendering image at a rendering device requires this inversion step, i.e.. finding the colorant mixtures to realize a desired color which may be given in device independent coordinates or in the color coordinate system of some other device (such as the CMYK monitor mentioned above.) Due to the complexity of the problem. it is generally not feasible to perform forward model inversion in real time as imagery is rendered. Practical approaches rely on interpolation in a sparse matrix of inverse function values. However. offering interactive control over key aspects of the separation transfonnation (aiso called SEP) implies that at least some parts of the calculation occur nearly in real time.

Because inversion of the forward model involves evaluating it potentially numerous times, acceptable performance requires a very rapid means of evaluating the forward model. One method involves de-l.-- -i of specialized hardware for polynomial evaluation. When specialized hardware is unavailable, software performance can be significantly enhanced via the same strategy as used for speeding rendering transformations: interpolate in a sparse table of forward model values, in particular in the FMT.

The proto SEP represents a color to colorant transformation with color coordinates as inputs and inkings as outputs. For illustration purposes, we will consider the case in which color coordinates are expressed in CIELAB units, U, a and 9. In order to build a sparse table of inverse function values for the proto SEP, the following addressing scheme is defined. If address variables, or inputs, are represented at 8-bit resolution. the span of the table in each dimension is 0 - 255 (2') digital levels. Continuous perceptual dimensions are mapped onto those levels so as to satisfy two criteria: a) the entirety of the input and output gamuts of interest must be represented, and b) adjacent digital levels are not perceptually distinguishable. Input and output gamuts will be detailed later; briefly. an input gamut is that of a device or application which supplies images (in this case for separation) and an output gamut is that of a device or application which receives the image in this case for rendering. With the above addressing scheme. a prototype separation table is computed. The following outlines the general process for building the proto SEP table.

For each grid point. or address, of the table, the amounts of the colorants needed to produce the input color are found. There are N tables, one for each of N colorants. assuming frame interleaved formatting. First. a starting point of the correct inking is speculated. Calculate its color by the forward model and compare the calculated color to the desired, address color. Save the color error. modify the inking, recalculate the color and see if the error is reduced. The forward model is also employed to calculate the partial derivatives of color with respect to ink. The resulting matrix, shown in the equation below, vields the direction of the movement in color occasioned by the change in colorants. The partials of color error with respect to each of the colorants indicate whether movement is in the correct direction. If negative. this indicates movement in the correct direction along that ink zero. then store each of the inkings in their respective tables.

da' aa"IaC, ai/aM, adlaY, c3i/c3K, dC db" = ',ab1aC, c11;/(3M, ab/(3Y, ab/(3K, - dM dL'.:aLIcIC,aL-laM,81:IaY,al:1(3K: dY dK The foregoing paragraph gives a simplified explanation of the algorithm embodied in procedures which use Newton's Method for finding roots ("Newton-Raphson") or one of many optimization methods for searching out "best" solutions in linear (e.g., the Simplex method of linear programming) or non-linear multidimensional spaces. The algorithm can also be implemented with neural networks, fuzzy logic or with the aid of content addressable memory, as detailed in prior art by Holub and Rose, cited earlier. A neural network can be trained on the results of repeated evaluations of a mathematical model. Further, neural networks may also be utilized to perform any of the color transformations of Figure 5. e.g.. building the forward model or rendering transform table. Color transformation with neural

C.

networks is described, for example, in U.S. Patent No. 5,200,816. cited earlier. Newton's method is applied to find the root of an error function, i.e., the solution is at the point at which the derivative of the error function goes to zero. Strictly, Newton's method is only applicable in situations where a solution is known to exist. Therefore, although it is an efficient procedure, it is supplemented by other optimization methods, such as "Simulated Annealing."

The model inversion technique described above is similar to those in Press. et al. sections 9.6, 10.8 and 10.9, respectively, already cited.) Optimization methods can produce a useful result when no exact solution exists. as in the case when the desired color is unprintable by (or out of the gamut of) the device. The optimization is driven by an error function which is minimized. typically by using downhill.

or gradient search procedures. Nevaen's Method and most of the 1A _'_ teehniques empiey the inverse of the jarebian matrix sheAm abev. In the case of Newton's Method. one of the colorant variables (in the four colorant case) must be fixed in order to find an exact solution, otherwise, there are infinitely many solutions. Usually, black is fixed. The algorithm for model inversion uses supporting search procedures in case the primary technique fails to converge to a solution.

The foregoing. simplified account overlooks the possibility that the gradient surface on the way to the solution has local minima in it which thwart the search procedure. This risk is minimized by the present inventors' technique of using the FMT to provide starting points which are very close to the solution. The above process for building the proto SEP Table is applied in system 100 as shown in the flow chart of step 6 in Figure 10A. For each color entry in the FMT, find the closest color address in the prototype separation table (step I 11). Then use the colorant address of the forward model table as the starting point in a search for a more exact ink solution at that color address of the proto SEP (step 112). The search and the addressing of the interpolation table can be in the cylindrical coordinates of hue. chroma and lightness or in the Cartesian coordinates L', a' and F. In the preferred embodiment, the forward model table has -at most 17X I 7X I 7X 17 - 84K grid points and the prototype separation table has at most 33X33X33 - 35K grid points many of which will not be within the gamut of the printer and some of which may not be physically realizable, i.e.. within human perceptual gamut. (This is the case because the gamut of a set of colorants is not likely to have a cuboidal shape suitable for representation in computer rn emory when it is converted from device coordinates to the coordinates of the color addressing scheme, as. illustrated in Figure 10B.) Therefore, there is a favorable ratio of starting points to solutions.

An important advantage of keying off the FMT is that most searches are guaranteed to be for printable colors. For colors near the gamut surface, Newton Raphson may fail because there is not an exact solution such that an optimization procedure is required. The search routine may use either the interpolative approximation to the polynomial (the FMT) for speed in a software-bound system or the full-blown polynomial for greater precision. In either case. the derivatives are easily calculable and the iterative searching procedures driven by gradients of color error (step 113).

The result of step 6 is a prototype color to colorant transformation (proto SEP) table in which virtually all printable addresses contain one or more solutions. Because of the favorable ratio of starting points to required solutions, many addresses will have solutions based on differing amounts of the neutral (black) colorant. The multiple black solutions are very useful in preparing to perform Gray Component Replacement and are stored in linked lists at the appropriate addresses of the proto SEP. which is stored in the local portion of the 1 VP (step 114). At this stage there are two ingredients of a final rendering transformation in need of refinement: the prototype gamut descriptor and the proto SEP.

An example of the results of step 6 is shown in Figure 10B, in which coordinates which are plotted within a cuboidal structure suitable for representing the perceptually uniform color space coordinates. Addresses having data are shown by black crosses. If the cube were subdivided into numerous cuboids so as to create an interpolation table, as described in Figure 9D, it is clear that many of the cuboids would not correspond to colors that are realizable on the device. The coordinates of the perceptually uniform space are L'. u and v from the CIE's CIELUV color space in the example.

After step 6, the prototype GD data of step 5 is refined into finished GD data at step 7 of Figure 5. Step 7 processes are shown in the flow chart of Figure 11. System 100 needs a favorable ratio of FN4T entries to proto GD addresses; therefore, most of the addresses get one or more solutions. Relatively few of these are surface Chroma points. The objective of step 7 is to use proto GD data as starting points for an iterative motion toward the gamut boundary using search and model inversion techniques described previously. Gamut surfaces often manifest cusps, points and concave outward surfaces- Therefore, quantization of the descriptor must be atfairly high resolution (128XI28 or more is preferred) and initiation of searches for a surface point from other surface points of different hue angle is often unfruitful. Because it is known what colorant mix (i.e., little or no colorant) is appropriate at the highlight white point of the gamut, it is best to begin refinement there, using solutions at higher luminances as aids at lower luminances (step 121).

At least one of the colorants must be zero at the gamut surface. Therefore. the strategy is to start near the desired hue angle from well within the gamut, move onto the desired hue angle and then drive outward until one of the non-neutral colorants goes to zero (step 123). It is often possible to zero the colorant that is expected to go to zero to hasten search pro,, iures (such as Newton Raphson) which require constrained iterations. Starting points are usually available from a linked list stored in the proto GD. that failing. one may be had from a neighboring hue angle. lightness cell, or. that failing, by setting out from neutral in the desired hue direction (step 122). If polvnomial evaluation hardware is available. it can accelerate on-the-fly calculation of surface points by forward model evaluation. Given a hue angle and lightness, the device coordinates bounding a small patch of the gamut surface are identified and colorant mixtures within the patch processed through the forward model until the maximum chroma at the huellightness coordinate is produced. (step 124). In situations where it is necessary to fall back to methods involving model inversion. the hardware assist continues to be valuable because searching involves numerous evaluations of the polynomial and its derivatives, which are also polynomials. Once a refined gamut descriptor has been completed, it is written to the shareable portion of the VP (step 125).

Referring to Figure 12, a flow chart of the processes in Step 8 of Figure 5 is shown, filling in any holes which may exist in the Prototype Transformation and computing functions that summarize the trade-off (Gray Component Replacement. GCR) between neutral and nonneutral colorants at each color address. A "hole" is a color address which is printable but for which no inking solution has yet been found. The resulting proto SEP table is called a generalized color-to-colorant table because it does not provide solutions for a specific amount of black. Step 8 has two goals. First, it seeks solutions for any colors interior to the gamut which have none (step 131) and writes NULL into all unprintable addresses. The mapping of unprintable colors to within gamut occurs later in step 9 in response to gamut configuration data. which the user may select using the GUI screen of Figure 21F, which is described later. Second, it stores at each address the means to exchange neutral for non-neutral colorants (Gray Component Replacement.) The preferred process is to store several solutions which can be interpolated amongst using LaGrangian, Spline or generic polynomial interpolating functions (step 132). The exact specification of black utilization is incorporated later in step 9. which also may be selected by the user via the GUI screen of Figure 21E, which is also described later. The finished prototype color to colorant (protoSEP) transformation table is written to the shareable portion of the VP (step 133).

Step 9 of Figure 5 includes the following processes:

1) Converting colorants of the proto-SEP transformation table based on black utilization information specified in the black color data:

2) Building Color to Color' Transform Table based on gamut configuration data. e.g., gamut scaling. neutral definition or gamut filter(s): and 1 1 3) Combining Color to Color' Transform Table with a n-Ansformation table containing specific GCR solutions to provide a rendering table. The black color data and gamut configuration data may be set to defaults or selectedby the user as color preferences. as described in more detail later.

Referring to Figure 13, the processes of step 9 are flow charted. Although step 9 is operative of four-colorant rendering devices, morethan-four colorant devices may also be provided with a rendering table as will be explained in discussion of Figures 16A and 16B. First, the finished proto SEP table (generalized color-to-colorant) and the black utilization data are read (step 140). The latter include Gray Component Replacement, % Under Color Removal (%UCR or UCR., a term referring to a limitation on total colorant application or Total Area Coverage, TAC) and possible constraints on the maximum amount of black or neutral colorant to be used; all three together are referred to herein as "black utilization."

The prescriptions for black utilization come from one or more of the following sources: a) local or system-wide defaults, b) broadcast of custom functions and limit values from a "boss node" configured in network 11 and c) values defined through a user interface which may be applied locally or transmitted and hared. Note that -modifications of black utilization do not have colorimetric effects. For example, compensating increases in nea tral colorant by decreases in non-neutral colorants in GCR does not change the color noticeably. Fields in the VP control selection of the source of prescriptions for black utilization under particular rendering circumstances. A user may select the black utilization in the Graphic User Interface of the model software, which is described later in connection with Figure 21E. The black utilization data, which includes % UCR, maximum black, and the GCR function or black solution, are stored in the shared part of the VP.

The first step in preparing a rendering transformation. then, is to convert the data on GCR stored at each printable address into a particular black solution by referring to the curve giving %GCR as a function of density while observing the maximum black constraint (step 149). Thus, the entries in the proto-SEP transformation table are converted based on a specific GCR solution within a maximum black limit. The converted proto-SEP table is stored in the local part of the VP. The second step is to find the maximum neutral density at which the total area coverage limitation is satisfied. i.e.. %UCR (step 150). This is done b\ moving up the neutral, or CIE Lightness, axis through the quantization scheme from the minimum printable Lightness, examining colorant solutions, until one is found that is within the limit. The minimum Lightness so identified is stored for use in the gamut scaling process to be discussed presently. Although it is conceivable to store min Lightness values as a function of color, rather than simply as a function of Lightness, for purposes of gamut scaling, this is usually an unwarranted complexity.

* In order to convert the GCR-specific result of the foregoing methodology into a rendering transformation, it must be "married" with a "conditioning" transformation. The latter is a color-to-color' conversion expressible as an interpolation table of the format presented earlier. (Note that any transformation that is evaluated by interpolation and that can be fitted with acceptable color accuracy by a polynomial may be performed by suitable hardware for polynomial evaluation.) It serves the multiple purposes of scaling unprintable colors of the addressing scheme onto printable values, "aliasing" the color definition of neutral to accommodate user requirements (Figure 21E, 270) and effecting conversions among color addressing variables.

An example of the latter is the conversion from Cartesian CIELAB addressing to "Calibrated RGB" addressing which is useful if the image data that will be processed through a rendering transformation are represented as RGB. It will be described later that Conditioning Transformations (CTs) play an important role in verification and device control. A CT is often the result of the "concatenation" of several individual, conditioning transformations serving the purposes just identified. The applications of transform concatenation include: 1) the method whereby a separation table with many NULL entries is made useful for rendering by concatenation with a conditioning transform which performs gamut scaling with considerable savings in transform-generation time and 2) feedback control of the color transformation process as will be detailed later.

In c- -:r to minimize cumulative interpolation errors, intermediate color to color' transforms may be stored and/or evaluated at high precision. Wherever possible. exact conversions of table entries should be utilized. For example, once all the mappings of CIELAB color to CIELAB color' have been compiled. the conditioning table entries for a color coordinate conversion from CIELAB to "calibrated RGB," for example, may be computed using analytical expressions.

Note that system 100 does not preclude importation (as described in discussion of User Interface) of standardly formatted color transformations ("profiles") prepared by applications other than the Virtual Proofing application. Accordingly, the rendering transformation may incorporate user preferences indicated through a transformationediting tool (called a "profile editor" by some Application Software Packages.) A key purpose of the conditioning transformation is gamut scaling, described below. it is important to elaborate this point: If the target device for the rendering transformation is a proofer. then the output gamut to which colors are scaled may be that of its client. The relevant. conditioning data for all devices on the network resides in the Virtual Proof. If the client's gramut fits entirely within the proofer's. then substitution of client for proofer gamuts is used to render a representation of how imagery will appear on the client. The display and negotiation of compromises to be made when the client's gamut does not fit entirely within the proofer's is discussed in conjunction with Figure 21F, below. Note that, in addition to accommodating a client's gamut, successful proofing may require other mappings. For instance, tone scale remappings to compensate for differences in overall luminance levels between video displays and reflection media may be performed. Likewise, "chromatic adaptation" transformations to compensate for changes in illumination, viewing conditions. etc. may also be implemented by means of data structures or tables of what are called "conditioning" or colorto-color transforms (output color transforms) herein.

Elaboration of the concept of color "aliasing," presented above is also appropriate: Neutrals are conventionally defined in terms of colorant in industry; elements of the default neutral scale which appears in Figure 21E (step 270) generally do not have common chromaticity coordinates up and down the Lightness axis. In order to map colorimetrically neutral addresses onto the desired mixtures of colorants, it is necessary to convert the colorimetric addresses to the colors of the "colorant neutrals" - a process herein referred to as 11 aliasing" because one color address masquerades as another. The term has a different usage. familiar to signal processing. when used herein in connection with imaging colorimetry Specifically, the processes involved in making a color-to-color' transform (numbered I to 4 in Figure 13) are as follows: (Recall that the goal is to be sure that only printable address colors are applied to the GCR-specific SEP table in the course of making a Rendering Table. Normally, color out = color in at the outset. More detailed discussion of input gamuts, output gamuts and gamut -operators follows. The sequence of the processes is important.) 1) Negotiate garnuts: In preparing a rendering transform to enable a proofing device to represent a printing press the color addressing of the color-to-color' table may be defined by the input gamut - in terms of the color image data. The range of colors represented in the table is limited by the extreme image values in the three dimensions of Uniform Color Coordinates. Because thequantization scheme is regular (cuboidal) there will be color addresses which do not occur in the image data (recall Figure 10B.) These may be ignored.

or mapped approximately onto the surface of the image s gamut. In place of the image's gamut. a general representation of the gamut of the original medium of the image may be used.

This method is preferred, for instance, to Using the gamut of the press as the input gamut because so doing would require conversion of bulky image data into coordinates that are all within the press's gamut. An objective of system 100 is to provide means of interpreting image data in various ways without creating a multiplicity of versions of that data. An exception would occur if image data specifically from a press sheet were required to evaluate image structure due to screening, etc. as part of the proofing process.

The output gamut may be the lesser of the client's (printing press) AND proofer's gamuts. where "AND" connotes the Boolean operator whose output is the "least common gamut." This may be derived using gamut filters. as discussed later. In this negotiation. the proofer's gamut may be constrained to be that available within the Total Area Coverage (%UCR) limitation applicable to the press at the user's discretion. Other interactive tools ma". be used to control the definition of the output gamut subject to the ultimate restriction of what can be rendered on the proofing device within the default or selected %UCR constraint. if applicable. (Of course a video display proofers gamut is not intrinsically subject to a UCR constraint.) 1 51 - 2) Perform gamut scaling: Using gamut operators to be discussed later, map color values to color' and store the color' values at the color addresses in the table. 3) Perform neutral aliasing: In each lightness plane, the color of the conventionally neutral inking (Figure 21E, 270) is offset from the neutra.1 color coordinate a' = 1; = 0 by an amount a, b'. In order to map image neutrals to this offset color, the color' values in the conditioning table should be shifted in such a way that an address of 0.0 maps to a color' value of a'.b. The function which performs the shift may be designed so that the amount of the shift decreases with distance from neutral. 4) Transform Color Coordinates (if necessary): The reason for this and a method of implementation were suggested previously, namely, it is preferred to perform gamut operations in Uniform Color Coordinates. but the image data may be represented in a color notation such as "calibrated RGB" so that a rendering table must be addressable by RGB coordinates. Because exact mathematical equations generally govern the relationship between Uniform CIE color and calibrated RGB, each color' value in the conditioning table is converted to RGB by means of the equations.

The Color-Color' Transform (XForm) is produced by the above steps and then. concatenated with the GCR-specific SEP table. The result is a rendering table for transforming the color of input color image data into colorant data for the rendering device. which is stored in the local part of the VP.

The following discussion is related to providing gamut mapping data of the gamut configuration data. Color addresses that are not printable must be mapped onto printable ones. In the present application, the output gamut is that of the proofer of interest or the printer it represents. However, the input gamut is not fixed by the architecture or software design because it may vary and have a profound effect on the rendering of out-of-gamut and nearly-out-of-gamut colors. This is the case because the outermost, or limiting, colors varv greatly among possible input gamuts. Input gamuts that warrant distinctive processing include:

I Other proofing devices include both hardcopy and video display devices (VDDs). Hardcopy devices are likely to have gamuts that are fairly similar. requiring only minor relative adjustments. Self-luminescent. additive-color devices such as video displays have very differently shaped gamuts from reflection devices so that the best mapping from input to output as a function of color will be unlike that for a reflection device. Computergenerated images originate in applications which are likely to exploit the gamut of a video device. Many retouching and page assembly applications use the RGB coordinate system of the monitor to store and manipulate images because it facilitates display on the VDD and interactivity. In this case, the input gamut is often that of the VDD, even if the image was originally scanned in from film.

2) Printing presses will usually have smaller gamuts than the proofing devices that represent them, restricting what is to be used of the available proofing gamut. If printed images are captured by an imaging colorimeter as part of calibration or verification so as to constitute the image data,. the input gamut may be better approximated by the rendering gamut of the printer than the receptive gamut of the colorimeter.

3) Electronic or digital cameras will usually have much greater gamuts than any output device, necessitating a very significant restriction on what portions of the input gamut can be printed. Note, however, that the maximum gamut of a linear camera encompasses many colors that are not commonly found in natural photographic scenes. Accordingly,,it may be preferable to design mapping functions for this class of device that are based on the scenery as well as ones based on device capabilities.

4) In conventional photography, the scene is first captured on film before being captured digitally. The relevant input gamut is the rendering gamut of film.

5) Regardless of the input medium and its gamut, there may be imagespecific imperatives. For instance, a very "high-key" image, consisting almost entirely of highlights. lace, pastels, etc. may not make extensive use of the available gamut of a color reversal film. The best gamut mapping for this particular image is not the generic one for film.

The gamut mapping data is provided by a gamut operator which is a function which maps an input color to an output color. The process of constructing the gamut operator is shown in Figure 14. It is a common practice to "clip" the input gamut to the output. In other words. all colors outside the output gamut are mapped onto its surface. This may be done in such a way as to preserve hue, saturation (Chroma) or Lightness or a weighted combination of the three. System 100 can work with such operators and supports access to 1 them through the "Rendering Intents" function in the GUI, as shown latter in Figure 21F. However, invertibility, reciprocality and smoothness are preferred properties of gamut operators, especially when processing and transforming image data.

Invertibility is an important property of the function because it insures that no information is lost except that due to quantization error. Reciprocality means that a mapping of input color to output color may involve either a compression or reciprocal expansion of the gamut, depending on which gamut is larger. Smoothness, or continuity of the first derivative. reduces the risk of noticeable transitions ("jaggies") in images which are due to gamut scaling. A simple exemplar of an invertible, reciprocal operator is illustrated in Figure 14. It is presented to demonstrate and clarify key concepts and then an operator which is also smooth is explained. A mapping.of Psychometric Lightness, L', is shown, however, the same operator is applicable to CIE Chroma, C. as well. It is assumed that hue is to be preserved in the mapping., a separate tool is supported within the GUI to gamut operations later shown in Figure 21F, for situations in which users want to modify hues. Figure 14 depicts the two cases of reciprocality, one in which the dynamic range of the input device must be compressed in order to fit the output gamut and ihe other in which-input lightnesses can be expanded to fill the larger dynamic range of the output device. The operator introduced below is able to handle both cases without explicit user intervention, although operator override is also supported through the GUI of Figure 21F.

Define Lplvol as the greater ("lighter") of the minimum input and output L' values.

Lpvol = max(I,,, - rn, where the minimum input lightness may, for example, be the L value of the darkest color that can be reproduced on a positive reversal film and the minimum output lightness the L value of the darkest color printable on a reflection medium. L,,.0, is denoted in Figure 14A by a plain dashed line.

For iightnesses higher than Lc,ip, the gamut operator maps input to output lightnesses identically as indicated by the equation in the open space below the maximum L of 100. cushion" is put between Lplvol and Lc,p in order to insure that the mapping is invertible.

Lci, = Lp,,,, + (Lma, - cushion.

0.1 is a reasonable value for cushion, chosen so as to reduce the risk of losing information due to quantization error to an acceptable level. In the limit in which cushion = 1, the entire range of input L values is scaled uniformly, or linearly, onto the range of output U values.

In either case 1 or 2, all lightnesses between Lh. and L.,. -,, are scaled onto the range of lightnesses between L,,,, and L,,, -, whether the scaling represents a compression or an expansion. A piecewise-linear scaling function is illustrated below for simplicity. Note that all L values refer to CIE Psychometric Lightness in these equations whether they appear with a or not.

If ( L.. > - Llip) then L. Oul = Lj,, Else [(Le The concept can be extended by adding the property of smoothness described earlier. The operator is based on a continuously differentiable function such as sine on the interval 0 to n/2, which generally resembles the piccewise linear function described above in shape but has no slope discontinuity. A table (Table 1) of values of the function is given belo,, %,; the first column is a series of angles in radians, X, from 0 to n/2 (90'), the second, the sine of the angle, Y, and the third the fraction WX. If we set Y/X = (1 cushion), we can control the "hardness" or abruptness of the gamut-mapping implemented by the operator stated below the table in- for the case of cushion = 0. 1. For speed, the various evaluations implied may be implemented by interpolation in look-up tables. The operator described does not enable a purely proportional scaling (cushion = L) The latter is not generally desirable but is available to users through the gamut options of the GUI in Figure 21F.

TABLEI

Angle.x (rad) - sin(x) sin(x)/x 0.0000 0.0000 0.0873 0.0872 0.9987 0.1745 0.1737 0.9949 0.2618 0.2588 0.9886 1 0.3491 0.3420 0.9798 0.4363 0.4226 0.9686 0.5236 0.5000 0.9549 -6109 0.5736 0.9390 0.6981 0.6428 0.9207 0.7854 0.7071 0.9003 <- "cushion" a 0.10 0.8727 0.7660 0.8778 0.9599 0.8191 0.8533 1.0472 0.8660 0.8270 1.1344 0.9063 0.7989 If ( L',,,, < 707 ( (100 - L.,j 1 (100 - L,.,,,_.j sin [.785 (100 - L,j 1 (100 -1...

Else -- 785 (100 - L.,j / (100 - L.... arcsin.707 (100 - L,,,) / (100 Lmtn_ln) The operators presented above rely on gamut descriptors to find the limiting colors of the input gamut that must be mapped into the output gamut. Once corresponding surface points in the two gamuts have been identified, the scaling function is used to prepare the conditioning transformation.

In summan,, input garnuts can be very different from output gamuts. They often have a larger range of luminances (or dynamic range) such that it is preferable to perform a scaling in Lightness before working on Chroma or saturation. Secondly. scaling of Chroma is performed along lines of constant hue. Compensation for imperfections in the CIE models (embodied in CIELAB and CIELUV) of hue constancy or for elective changes in hue are handled separately. An example of elective changes is the need to map highly saturated yellows from film input to highly saturated printyellows by way of a change of hue angle. Modifications of hue can be accomplished through the conditioning transformation by color allasing.

Referring now to Figures 15A and 15B. the shareable and local components described above in the VP are shown. In the interest of compact messages. not all shareable data need be transmitted in each transaction involving Virtual Proof. To the application software which administers the Graphical User Interface Software. both operating at a node, the VP is a set of data structures based around the classes Node and Devi c e/Trans form. The data structures map onto a general file structure which is not bound to a particular operating system. computer platform or processor. Object-oriented conventions of data-hiding are employed in the software to insure the integrity of transformations which are manipulated at a node. Accordingly, the VP flies which store the data and transformations have local and shared components, as stated earlier; shared components consist of data which are read-only to all nodes except the one assigned responsibility for the data. During initialization in preparation for virtual proofing described in connection with Figure 18, participating nodes insure that appropriate data are written to the relevant nodal fields within the shared file system.

The VP enables revision of color aspects of page/image data up to and even during 0 rendering. An important aspect of revisability is the custornization of data for rendering on particular devices. An equally important properry of revisability is that page/image data need not be altered directly: rather they are "interpreted" in various ways through the medium of the VP. This eliminates the need to maintain multiple, large versions of page/image data at a particular node or to move one or more versions around the network repeatedly as a result of re-interpretation. Therefore, although the VP allows for image specificity and linking, preferably it is not bound into page/image files. The structure of the VP is similar to that the Tagged Image File Format (an example is described in "TIF I F TM Revision 6.0". June 3. 1992, Aldus Corp.. Seattle WA, pp. 13-16).

An example of the tagged or linked list file format for the shared parts of the VP is shown in Figure 15C. The tags are referred to as Field IDs (bottom right of Figure 15C.) It is possible for a given device to be present in a VP file more than once, specialized to represent images for different input gamuts or black utilization, etc.

System 100 may incorporate rendering devices at nodes having more than four colorants. The processes for performing color to colorant conversions for more than four colorants is shown in Figures 16A and 16B. which are connected at circled letters A and B. In Figure 16A. after starting, if the extra colorant is neutral. then proceeding continues to the start of Figure 16B. Otherwise the followincy steps for adding additional non-neutral colorants cr (e.g.. Red. Green and Blue) are performed:

Step 1) Proceed as for 4 inks through to the stage of gamut scaling. Use the Black Utilization tool which enables %GCR to depend on Chrorna, to push towards maximum Black 1 1 I - 57- (2-colors + Black, with the complementary color pushed toward zero) solutions. Save this as an intermediate table. This intermediate is the equivalent of a GCR-specific SEP table.

Step 2) Build a model for converting C, M Blue and K (black Or N) to color and omitting the colorant complementary to the "auxiliary," in this case, Yellow. Make a Forward ( Z+eo I) Model Table and use the model to extend the original gamut descriptor prepared inA+a+. Do likewise for C, Y, Green and K and M, Y, Red and K. Note that the general rule is to add additional colorants one at a time, grouping each with the colorants which flank it in hue angle. Make FMTs for each new model for each auxiliary colorant and re-refine the Gamut Descriptor. Note, however, that the multiple models are used to refine only one GD.

Step 33) Modify the proto-rendering table (only one of these is maintained): Within the C.M.Blue.K gamut. there are not multiple solutions with different amounts of black at a given color; however, there are many solutions trading off CM vs. Blue. Store linked lists of these solutions at relevant color addresses in the intermediate table. Do likewise for CYGreenK and MYRed.K.

Step 4) Treat the intermediate table as a "prototype table" in the 4colorant case. Perfect it by making sure that all printable addresses in the new color regions of the table have at least one solution ("fill the holes.") Step 5) Once the intermediate table has been reprocessed for all non- neutral auxiliary colorants, convert to a rendering table by performing the analog of GCR in the three new regions of the table. Check total area coverage. re-scale gamut and complete as for four inks (Step 9, Figure 5.) The foregoing procedure does not estimate the full gamut available; for example. the gamut at the hue angle of cyan is increased by the availability of blue and green. In other words, the BCGN gamut (where N stands for Neutral, or black) is not considered in the foregoing. Overprints of Blue and Green are likely to be substantially dark, compared to cyan. There: the additional colors available in this gamut are not. in general. very numerous and need not be computed. However. in those cases in which the lost colors are impor-tant. the procedure outlined above is extended to include auxiliary gamuts centered on the standard subtracilve primaries (C. M and Y) rather than on the additional colorants (R. G and B.) The result is overlapping auxiliary gamuts. By default, the decision regarding which 58 - of the overlapping auxiliary gamuts to search for a solution chooses the gamut centered on the ink of closest hue angle. There is nothing in the procedure which prevents its extension to even more colorants, which may be added in a recursive fashion. However, practical applications involving the overprinting of more than 7 (or 8, in the case of an extra neutral) colorants are very unlikely.

After the above steps are complete, if more colorants need to be added, processing branches to the circle A at the start of Figure 16A, otherwise the process for adding additional colorants is complete. In the case of addition of auxiliary colorants which does not involve overprinting more than 3 or 4 colorants at a time (as in the case of multiple, "custom" colorants that might be used in package printing) the colorants are treated as separate sets according the procedures outlined previously.

If the process branched to Figure 16B (to circle B), then the following steps for adding an approximately neutral colorant, such as gray. are performed: If additional nonneutral colorants are also to be added, add them according the procedure outlined in Figure 16A above.

Step a) Prepare a colorant to color transformation for the 5-colorant set CMYKGray. Evaluate this model either with polynomial hardware or with a 9X9X9X9X9 interpolation table having about 60,000 five-dimensional cells. The simple linear interpolator is preferred and is panicularly appropriate to this situation because the complexity of calculations scales linearly with the dimensionality of the interpolation space. As usual. make a Gamut Descriptor in tandem with building FMT.

Step b) Invert the model as in the 4-colorant case. fixing black and gray; build up linked lists of alternative solutions for a given color.

Step c) Proceed as in the 4-colorant case. When inverting the model. use only, CMY and Gray wherever possible (i.e., fix black at zero,) adding black only as necessary to achieve density.. There are two stages of GCR. In the first. black is held to a minimum and gray is exchanged with C. M and Y. In the second. black may be exchanged, optionally. with gray and small amounts of the other colorants. as needed to keep the color constant. In the second stage. an error minimization routine is needed: Newton-Raphson is not appropriate.

1 59 - Step d) UCR, preparation of a conditioning transformation, and so on as in the 4- colorant case, follow the second stage of GCK Complete therendering table, except for rA'+er +he above z:tcp-- a-d addition of auxiliary, non-neutral colorants. are t.,, be ddde, p c - LO:Z 1 01:7 Far A +re proc.-.-,: -Q. add" Referring to Figure 17, the process f6i building a gamut filter is shown. The finished a dd; +, prototype color to colorant table, filled with NULL cells and specific GCR solutions, is one cc to I' e" manifestation of a device's gamut. It can be converted into a very useful filter in which each cell gets an indicator bit, 0 if the cell is NULL and I otherwise. The filters of two or more devices can be used to enhance visualizations.

Many graphics cards for driving video displays provide an alpha channel or overlay plane enabling the visualization of translucent graphical information on top of a displayed image. The results of performing Boolean operations on combinations of gamut filters for multiple devices may be converted into color-coded or pseudocolor overlay information to reveal things such as which image pixels are in gamut for one device but not another. With this tool. the intersection of the gamuts (the "Least Common Gamut") of five presses can readily be compared with each press's gamut in turn on common imagery, in support of a decision about what to use as a common criterion tor color reproduction across the network.

Se.mi-transparent overlays are generally not possible for hard-copy devices without extensive image processing. In the case of a printer, a monochrome version of the image may be prepared and printed, overlaid with colored speckles indicating regions of overlap of multiple gamuts. The "overlay" is actually constituted by redefining subsets of the pixels in the image which belong to a certain gamut category as a particular speckle color. The foregoing method involves making a modified copy of the color image data with reference to the gamut filter.

An alternative, and preferred, method provided by the invention for readily identifying gamut overlaps is to filter the color-to-color' transform or actual rendering table. This may be done by leaving the contents of "in" addresses as they were while modifying the color or coloranE contents of one or more varieties of "out" addresses to contain white or black or other identifying color. Different identifying colors may code different regions of intersection. overlap or disjointedness of the gamuts of several devices. When one or more channels of the (original) color image data are rendered through the "Filtered rendering table."

60- colors in the image which are out of gamut are mapped to one of the identifying colors and the resulting print reveals gamut limitations of various devices with respect to the imagery itself. An additional advantage of this method is that it is effective even when some of the colors under consideration are out of gamut for local proofing devices.

Another method of visualization available with the filters is to look at slices through Boolean combinations of the filters for two or more devices. Finished proto SEP tables are not generally useful other than in the preparation of rendering tables for a particular device, therefore, they are kept local. The filters are written to the shared portion of the VP.

After the &beye steps-& d, if additional nen neutir-all celor-ants are alse to be ftddcc,- - 14 A branehes te eireled A in L pr-oraess-f-er adding additional eelerants, ends.

More specifically. in Figure 17 a flowchart of the process of making a gamut filter in either a compressed form. in which a single bit is 0 or I to denote NULL or "in-gamut," or in a form in which each color's status is coded with a byte equal to 0 or I is illustrated. In the compressed case, a computer word may be used to represent a row or line of the prototype color-to-colorant table with each bit packed in the word representing one color address. A.fier reading the proto-table (step 1,) the steps of making the compressed filter include 2c) looping over the lines of the table, 3c) setting or resetting the bits of a word depending on printability and 4) finally writing the filter to shareable VP. The steps for making a byte table are completely analogous, except that a byte is devoted to each color address.

The two basic processes of calibration and verification (monitoring) rely on instrumentation and interact to produce the rendering transform which embodies the VP and can interpret color image data.

Referring now to Figures 18A-B, a flow chart of the program operating at nodes 102 and 104 for virtual proofing in system 100 is shown. These figures are connected at circled letters A-D.. - the top of Figure 18A, one or more users invoke the application software 1801: a single user can run the program in order to revise the Virtual Proof as it relates to the local node. or to other nodes which are accessible. Often. multiple users run the program simultaneously to negotiate the VP. Network readiness is established by making sure that the relevant, participating nodes all share current data 1802. CMI"s are put through (preferably 1 automatic) calibration checks 1803. Next. verification of device calibration is attempted by rendering and analyzing color 1804. The details of this step depend on the nature of the device and of the color measurement instrument; if the device is a press, particularly one with fixed information on the plate, the verification is most likely to involve "live" imagery rather than a predefined cal/verification form or target such as one consisting of tint blocks. The color error data 1805 produced by verification are supplied to the press controls. if appropriate 1806, as well as to the program to support decisions about whether color variations can be compensated by modification of existing rendering transforms of the VP or whether recalibration of the device is required 1807.

If re-calibration is called for, the program branches to C. atop Figure 18B. where svstematic calibration after Fig. 5 is undertaken 1808. Else. the program branches to B. atop Fig. 18B, to revise the color-to-color' transform 1809 based on the processing of the color error data, which is detailed later in Figures 19 and 20. Nem the need for User-preference revisions is assessed at D. If YES, then gather User preference data 1810 and re-specify rendering transforms 1811 as in Step 9, Fig. 5. If NO, then revise VP for a new interpretation of image data and render 1812. If results are satisfactory, conclude. Else, either recalibrate at A or revise preferences at D. depending on the nature of the diaimostics.

Verification is a feature of system 100 used in virtual proofing. described above. and color quality control of production rendering devices. The reason for verification is that the use of system 100 for remote proofing and distributed control of color must engender confidence in users that a proof produced at one location looks substantially the same as one produced in another location, provided that the colors attainable by the devices are not very different. Once rendering devices are calibrated and such calibration is verified to each user aiiewing. virrual proofing can be performed by the users at the rendering devices. In production control, such verification provides the user reports as to status of the Color quality.

During verification of production rendering devices, on-press analysis of printed image area may be used in control of the production process and for accumulation of historical data on color reproduction performance. Historical data may be used in a statistical profile of the manufacturing run which serves as a means of verifying the device calibration. It is also used to inform and update the virtual proof, enabling better representation of the production equipment by a proofing device. With a sufficient accumulation of historical information, it is even possible to model and predict the effects of neighboring pages in a signature on the color in the page of interest to an advertiser.

Once a device has been calibrated. the color transformations thus can be one of the mechanisms of control. In a control loop, colors produced by the device are compared to desired values and mechanisms affecting colorant application are modulated to reduce the discrepancy between measured and desired values. Control implies continuous feedback and iterative adjustment over many printing impressions, whereas proofing devices are usually one off. Nevertheless, proofing devices vary with time and periodic recalibration is a means of control.

One feature of system 100 is to provide the User with information about the color accuracy of the proof It has been noted that the invention is compatible with two kinds of instrumentation, unitary colorimeters (SOMs 13) capable of measuring homogeneous samples and imaging colorimeters (imagicals 14) capable of sensing many pixels of complex image data simultaneously. In the following discussion, differences in verification procedures for the two kinds of instrument are considered.

Calibration is best carried out with a specialized form which is known to explore the entire gamut of the device. The rendered form can be measured by either type of instrument. In verification, the requirements of sampling the entire gamut are not as stringent; the interest is often in knowing how well the reproduction of all the colors in a particular image is performed, even if the image samples only a part of the device gamut.

Referring to Figure 19. steps 1804-1807 of Figure 18A are shown. A specialized verification image is analyzed either with a SOM 13 or an imagical 14 according to the following procedures: Step 1: Render an image consisting of homogeneous samples ("patches") of three types: a) patches whose nominal colorant specifications match those of patches in the original calibration. b) patches whose nominal specs are different and c) patches specified as color. The original calibration procedure (see Fia. 8) produced statistical estimates of within-sheet and between- sheet color differences or process variation. User-defined requirements for accuracy are expressed in terms of standard deviations (or like

1 1 63 - quantity) within the process variation to define confidence limits for the process. Three kinds of color error derived from verification procedures are used to control the process and are referred to the confidence interval in order to decide if recalibration is necessary.

Step 2: New measurements of type "a" patches (preceding paragraph) are compared to historical values to estimate change in the process, a thorough sampling of color space is useful because it is possible that change is not uniform with color. Step 3: Model predictions of the colors of patches of types "a" and "b" are compared to measurements to yield estimates of the maximum and average values of color error for the forward model. Step 4: Comparisons of the requested colors of patches of type "c" to those obtained (measured) are used to estimate the overall error (due to change of the process, model error, model inversion error, interpolation and quantization errors, when relevant.) Step 5: If color errors assessed in this way exceed the confidence limits, the User(s) are advised that the system needs recalibration and corrective actions may also be taken, such as modification of conditioning transforms, depending on the severity of the problem. If the device is a press, color error data is furnished to the press control system (subject to User intervention) which does its best to bring the press as close to criterion as possible.

The advantage of specialized imagery is that suitably chosen patches provide more information d= may be available from imaging colorimetry of images with arbitrary color content. This can guarantee that the entire gamut is adequately sampled and can provide differentiated information about sources of error. Also. imaging colorimetry of the reproduction during or shortly after rendering is often the least obtrusive way of verifying that forward models and rendering transformations are current. especially when a volume production device is the object.

Because, as was noted above, the variations in color need not be uniform throuLhout the gamut of the device. the data structure is segmented into clusters of contiguous cells in order to identify the most frequent colors in the various regions of the gamut. Thus. system herein samples color errors throughout the image peints. The processin.0 checks to make sure that the frequency it is reporting for a cluster of cells is a peak. not a slope from a neighboring cluster.

In order to improve the reliability with which corresponding peaks in different histograms can be resolved, methods of image processing such as accumulation of counts from multiple images (averagingJ bandpass filtering and thresholding are employed. Then regions of the histograms are cross-correlated. Cross-correlation is a technique discussed in many texts of signal and image processing, in which two functions are convolved without reflection of one of the two. It is similar to techniques in the literature of W. K. Pratt, Digital Image Processing, NY: Wiley, 1978, ch. 19, pp. 551-558. A "cross-correlogram" reveals the offsets of one histogram with respect to another in three-space.

The color offsets of the peaks are expressed as color errors. These are made available for numerical printout as well as for visualization. In the latter, the user may choose to view a monochrome version of the image overlaid with translucent color vectors showing the direction and magnitude of color-errors, or may choose to view a simulation of the error in a split screen, full color rendering of two versions of the image data showing what the error can be expected to look like.

For clarity, an equivalent procedure for cross-correlation can be outlined as follows: 1) subdivide the histograms into blocks and "window" them appropriately, 2) calculate Fourier Transforms of the histograms, 3) multiply one by the complex conjugate of the other, 4) Inverse Fourier Transform the product from 3) and 5) locate the maximum value to find the shift in the subregion of color space represented by the block.

For the simplest level of control, the inverse of the color errors may be used to prepare a conditioning transformation which then modifies the rendering transformation employed in making another proof. For more sophisticated, on-line control, the data are used to compute error gradients of the sort described earlier and used by optimization and error minimization algorithms. Results are fed to the control processor of a press or used to modify the rc:;: - P I& iT rendering transform as a control mechanism for ali;which does not use a press bearing fixed information.

The goal is to determine errors in color reproduction in a resolutionindependent way. This is shown in reference to Figure 20. illustrating processes 1804-1807 in Figure 18A when an imagical 14 is verifying using live image data. In step I of Figure 20. the histogram is defined. Generally. it is a data structure addressed similarly to the Conditioning Transform 1 (color-to-color' table) described earlier, although the range of colors represented in each dimension of the structure is adaptive and may depend on the imagery. In step 2, the 3-D arrays to hold accumulated histogram data are allocated in memory and initialized; one array is needed for "live" image data and the other for reference data. At step 3, capture of "live" image data occurs. Optical low-pass filtering may precede image capture, preferably by a solid state electronic sensor, in order to reduce aliasing in signal processing. The electronic. pixel data are convened into CIE coordinates.- and. simultaneously, a histogram of the relative frequency of occurrence of colors in the image is stored. As mentioned earlier. the data structure may be segmented into clusters of contiguous cells in order to identify the most frequent colors in the various regions of the gamut.

In pan 4 of the process, the image data (not that captured by the imagical. but the "original," image data) are processed through the colorto-color' transform to produce Reference Color Data which are accumulated -in -a _histogram structure. It is important to recognize what the requested (or "reference") colors are. They are the colors (preferably in CIE Uniform coordinates) which are the final outputs of all the color-tocolor' conditioning transforms (usually exclusive of color-coordinate transforms)'and thus represent the interpretation of the image data negotiated by the Users.

In steps 5 and 6, as described above, the program checks to make sure that the frequency it is reporting for a cluster of cells -is a peak, not a slope from a neighboring cluster. Accumulation of counts from multiple images (averaging,) bandpass filtering of histograms. thresholding, autocorrelation and other operations of image processing are used to improve reliability and the ability to resolve peaks and match corresponding peaks in different histoarams. Ordered lists of peaks are prepared and written to the shareable portion of the VP. The lists are compared and corresponding peaks identified. The color offsets of the peaks are expressed as color errors. In step-7.-coloT error data are made available to User/Operator and control systems.

Referring to Figures 21A-21F. a Graphical User Interface (GUI) to the application software is shown. The GUI is part of the software operating at the nodes in network I I for convevine the workings Of system 100 at ahigh-level to the user, The user-interface has reusable software components (i.e.. objects) that can be configured by users in a point-and- 66- click interface to suit their workflow using established visual programming techniques. The GUI has three functions: 1) Network (configure and access resources,) 2) Define (Transformation) and 7' Apply (Transformation.) All three intemet. For instance, verification furictions fit logically within Apply Transformation but must be able to feed back corrective prescriptions to Define Transformation which provides a superstructure for modules concerned with calibration and modelling. Therefore, both Define and Apply need access to Color Measurement modules, whether they.make use of imaging or non-imaging instruments. "Network" is responsible for coordinating network protocols and polling remote nodes. Part of this function includes the identification of color measurement capabilities of a node. Another part is to insure that a users mapping of software configuration onto his workflow is realizable. Loading the appropriate color measurement device drivers is as crucial as choosing and initializing the correct communications protocol and proofer device drivers. Therefore, Color Measurement coexists with modules for administering network protocols. building device models. building color transformations and implementing the transformations for the conversion of color images.

For the purposes of this discussion, assume that the application is stand alone. Today. Graphical User Interfaces can be prepared via automatic code generation based upon re-usable components in most of the windowing environments on the market. The depictions of menus and attributes of the User Interface in what follows are not meant to restrict the scope of the invention and are kept simple for clarity.

Referring to Figure 21A, a first level hierarchy screen is shown allowing a user to enable configuration of network 11 nodes, remote conferencing. and user oversight of processes in system 100. Upon invocation, the application introduces itself and offers a typical menu bar with five choices (command names), File. Network. Define. Apply and Help. Clicking File opens a pull-down menu whose selections are like those indicated by 221 in the figure. Clicking on Create 222 initializes file creation machinery and opens the Network tableau (Figure 21B) in a mode of readiness to design a virtual proofing neTwork. Export VP (Virtual Proof) 223 offers the option of converting color transformational components of the Virtual Proof into standardized file formats such as International Color Consortium Profiles. Adobe Photoshop color conversion tables, PostScript Color Rendering 1 Dictionaries, TIFF or TIFFIT. A possibility of importing standardized color transformation filcs is also provided. Other menu items under File are conventional.

The Network 1,i-tding 224 opens a tableau concerned with membership in the virtual proofing network, the physical and logical connections among nodes and the equipment at the nodes and its capabilities. The Define heading 225 provides means of calibrating (characterizing) devices, enabling customized assemblies of procedures to be applied to appropriate target combinations of devices and color measurement instrumentation. The Apply heading 226 covers display of imagery rendered through the virtual proof and provides tools for verifying and reporting on the accuracy of color transformations. customizing those transformations, conferencing, comparing various versions of the proof and establishing contact with other applications performing like functions. The Main menu offers Help and each of the Network, Define and Apply menus offer Tutorial interaction.

Clicking on Network in the Main menu opens a tableau of Figure 21B. which is concerned with Connections and Capabilities. Clicking Connection 227 reveals a sidebar concerned with tools and attributes of network I I connections. For example, during creation (see Figure 1,) it is possible to pick up a wire by Way of the "wiring" entry of the sidebar 228 and move it into the field as part of assembling a network model that can be written to a file and can direct proofing commerce. Double clicking on the wire reveals information about the connection or permits editing of the information when the "Modify" radio button 229 is activated. Error notices are generated when the software drivers required to implement the model are not available or when the hardware they require is not present. The present invention is not wedded to particular current (e. g.. modem, ISDN, TI, satellite, SMDS) or anticipated (ATM) telecommunications technologies, nor to particular networking protocols. Nodes can be defined, given addresses, security. etc. and can be equipped with proofing devices in a mamer similar to the design of connections. The summary of the network"s connections and of nodal capabilities is shared through the Virtual Proof s tagged file format described earlier which is kept current at all sites.

In the center of Figure 21B is an example network topology. It resembles the network of Figure 1. where "ch" 230 refers to a possible client (e.g., adverliser) member. 1. ad" 231 an ad agency, "pub" 232 a publisher, "eng" 233 an engraver and the Ps are printers.

C The links between the nodes are created and modified through the wiring 228 functionality of "Connection" mentioned above. Clicking "Capability" reveals a sidebar 234 concerned with devices and their associated instrumentation for color calibration and verification. An example of the use of the menu is as follows: I am a user at an ad agency who wants to establish a remote proofing conference regarding a page of advertising with a publisher and printer. I bring up the relevant network using "Modify..." in Figure 21A, push the radio button for view/select 235 in Figure 21B, and click on the "pub" node 232. This creates a connection between the ad and pub nodes if one can be made and initiates a process of updating virtual proof files at either end of the link. Then I click on hard proof 236 and color measure 237. This utilizes the updated VP information to show me what hard copy proofer(s) are available and how they are calibrated and/or verified. Then I follow a similar sequence of actions with respect to a P node. The initiation of display and conferencing about color image data is done via the Apply menu of Figure 21D.

To pursue the example of the preceding paragraph. suppose that I find that the hard copy proofer at node "pub" has not been calibrated recently. A study of the information about the device in the updated Virtual Proof reveals whether re-calibration or verification procedures can be carried out without operator intervention at that site. From one site or the other, the Define (Transformation) menu of Figure 21C provides access to the tools and procedures for calibrating while the Apply (Transformation) menu (Figure 21D) supports verification. A node can be activated in the Network menu and then a device at the node singled out for calibration within Define.

Clicking on "Node" 241 in the bar atop the Define menu of Figure 21C opens a pull down providing access to other nodes without requiring a return to the Network menu. Which node is active is indicated at the upper left of the menu 242. Clicking on "Devices" 243 opens a pull-down which lists classes of devices; clicking on a member of that list displays an inventory of devices of that class present at the active node. Selecting a device in this list is the first step in the process of calibration and causes the device to be identified as active 248 at the top of the menu. The classes of devices of particular interest in the invention are imaging colorimeters or imagicals 14 ("imagicals." 244.) unitary colorimeters (SOMs 133. capable of measuring discrete tint samples. 245j presses 246 and proofers 247 of hard and soft copy varieties. Clicking on "Procedures" 249 reveals a pull down listing calibration modules 250 such as linearization, Forward Model Generation, etc.

Procedures appropriate to the device can be dragged and dropped into the open field at the center of the menu and linked by connecting arrows. The controlling application software monitors the selections, performs errorchecking (with notification and invocation of tutorial material) and links together the modules needed to perform the task if possible. Figure 21C shows a flowchart for complete calibration of a rendering device encircled by a dotted line 251. In the case of some proofing devices, such as Cathode Ray Tube displays and Dye Sublimation printers, it may be sufficient to perform complete calibration only infrequently. In particular, it is usually adequate to re-compensate for the gamma function of a monitor (a process which falls under "linearization") on a regular basis. Because phosphor chromaticities change very gradually. redetermination of the color mixing functions of the calibration need not be performed as frequently. Therefore. a user may activate the CRT device at his node and specify only linearization in preparation for a proofing conference. Alternatively, the present invention covers the equipment of monitors with devices that can participate in continual verification and recalibration.

The "Apply" Transformation menu (Figure 21D) provides access to the database of pages and images that are the subject of remote proofing through the "Page/Image" 256 selection in the menu bar. Clicking here displays a shared file structure. Although the (generally) bulky color image data file of interest need not be present in local storage 19 of Figure 3A at all nodes where proofing is to occur, it is generally desirable to make a local copy so that rendering across the network is not necessary. However, one of the purposes of the Virtual Proof is to make multiple transfers of the bulky data unnecessary. The "Node" 257 and "Devices" 258 elements of the menu bar have effects that are entirely analogous to those in the "Define" menu of Figure 21C. More than one device at a node can be made active in support of a mode in which interactive annotation and conferencing x-la the medium of a soft proof on a video display is employed to negotiate changes in a hardcopy. remote proof that is taken to be representative of the ultimate client device.

Clicking on "Procedures" 259 in the Apply menu bar of Figure 21D reveals a pull down that includes functions such as "Render to displa 260, "Veri%- " 261 and "Window..." 262 to external applications. Rendering supports display. either within the Apply window or on a separate, dedicated video display, of imagery a) as the designer imagined it, to the extent that theproofer is capable of showing it, b) as a client device, such as a press, can reproduce it and c) as another proofer is capable of representing the press, including indications of -gamut mismatches superimposed on the imagery and location of errors identified by the verification process. To fin-ther the example, the virtual proof may mediate a rendering of an image as it appears or will appear on press. If the node includes an imaging colorimeter, then an image of the proof can be captured and analyzed in order to provide verification of how well the proofer represents the client. Without verification. digital proofing and remote proofing for color approval are not really useful.

The Apply menu provides a Window 262 through which to "plug-in" to or to "Xtend" applications such as Adobe Photoshop or Quark Xpress. It also provides the platform for incorporating remote, interactive annotation of the sort provided by Group Logic's imagexpo. reviewed earlier. Imagexpo focusses on marking up images with a virtual grease pencil, a concept that is extended in the present invention to remote conferencing concerning gamut scaling, black utilization and other aspects of the definition of rendering transforms. Aspects of rendering such as black utilization 263 (or gamut operations) can be harmonized across the production network by sharing/exchanging black designs through the virtual proof file structure and mechanism.

Menus supporting interactive design and selection of user preference data are shown in Figures 21E and 21F. A User-interface to support interactive determination of black utilization is depicted in Figure 21E. It may be invoked from either Define or Apply menus. At the top"T- 270 is shown a panel of colorant specifications which are to be considered neutral in color. Users may redefine entries in the panel by clicking and keying or b moclifying curves in the graph below 272 provided the graph is toggled to neutral definition rather than GCR. In the neutral definition mode the user may move points on any of the colorant functions. the points are splined together and changes in the graph are reciprocal with changes in the panel. A "retum to default" switch 274 provides an easy way to get out of trouble. At the upper right 276. 278. variable readout switches enable definition of maximum colorant coverage and maximum biack. At the bottom right 280. "Customize Tonal Transfer"

1 opens the door to user modification of one or more of the I -dimensional output postconditioning LUTs which are part of the color to colorant transformation. The application warns stemly that the specification of transfer curves which did not prevail during calibration will void the calibration; however. there are situations in which knowledgeable users can make effective use of the flexibility afforded by this function.

When the Graph is switched 282 to GCR mode, the user can control only the shape of the neutral colorant curve; because GCR is colorimetric, the nonneutral curves respond compensatorily to changes in black. The relationship of the curve specified in the graph to the amount of black chosen for a solution at a particular entry in the separation table is as follows: The function shown in the graph represents amounts of the colorants which produce given neutral densities. At each density, there is a range of possible black solutions from minimum to maximum. At the minimum, black is zero or one or more of the non-neutral colorants is maxed out; at the maximum, black is at its limit and/or one or more nonneutral colorants are at their minimum. In this invention, the %GCR is the percentage of the difference between min and max black chosen for a solution. By industry convention, a constant %GCR is seldom desired for all Lightness (density) levels. Therefore, the blac, curve in the graph defines the %GCR desired as a function of density. Although it is conceivable td make GCR a function of hue angle and Chroma as well as of Lightness. this is usually an unwarranted complexity with one exception: It is useful to graduate GCR with Chroina when preparing transformations for more-than-four colorants as discussed earlier. This need is addressed through the "GCR special..." submenu 284 offered at the bottom right of Figure 21E.

Figure 21F depicts a GUI screen to gamut operations. Clicking on "Gamuts" reveals a pull-down 286 which provides access to lists of input, output and client gamuts: the latter two are both output gamuts. but a client gamut is known to the system as one that is to be represented another device. It is possible to drag and drop members from the different types into the field of the menu and to link- them with various procedures. as was the case in Figure 21C. Procedures applicable to the gamuts include: 1) Analyze 288 which displays information on how a particular conditioning transformation was put together (e.g.. was it a concatenation of gamut scaling and color aliasing operations) - which ones?) and on gamut compression records, a constituent of the VP which stores key variables of a gamut scaling, such as minimum Lightness, cushion value, functional form. etc. 2) Apply to Imagery 290 enables display of imagery mediated by n-ansformations configured in this or related menus on some device. 3) Compare Gamuts 292 enables visualization, in several forms, of the relationship between the gamuts of two or more devices - this functionality is elaborated in a following paragraph. 4) Concatenate 294 does not apply solely to gamut operations; it links nested or sequential transformations into implicit, net transformations. 5) Gamut Operator 296 provides a graphical display of an operator; this is a different representation of the information available from Analyze 288. 6) Negotiate Proofing Relationship 298 works closely with Compare Gamuts 292; it enables users to make decisions based on information provided by Compare, such as whether to use the Least Common Gainut as the aim point for a network of volume production devices. 7) Remap Hues 300 provides the separable hue adjustment functionality described earlier. 8) Rendering intents 302 is a mechanism for providing users with generic gamut scaling options for accomplishing things like obtaining the most saturated rendering on paper of color originally created in a business graphics application on a video display. Compare Gamuts 292 allows a user to invoke and control-the use of gamut filters, which were described earlier in connection with Figure 17.

System 100 supports the coordination of color control in multisite production in several forms. Because the virtual proof encompasses records of the states of calibration of network devices independent of the color data, application software can define a criterion for reproduction across the network in one of several ways. Based on the states and capabilities of network devices, a criterion may be selected which all devices can satisfy, such as a least common gamut (LCG). LCG defines the aim point for production and the control system strives to minimize the color error with respect to it. Alternatively, users may select one device as the criterion and all others are driven by the controls to match it as closely as possible. Optionally. users may choose to disqualify one or more rendering devices on the network because it/they cannot match the criterion closely enough or due to failures of verification.

1 7 3 Although the above description relates to the printing and publishing industry. it is also applicable to other industries, such as textile printing. Further, in packaging and related industries. more than four colorants may be used in circumstances in which no more than four colorants overlap within a given region of the page. The system can accommodate this by applying the herein methods to separate regions of the page.

From the foregoing description, it will be apparent that there has been provided a system. method. and apparatus for distributing and controlling color reproduction at multiple sites. Variations and modifications in the herein described systems in accordance with the invention will undoubtedly suggest themselves to those skilled in the art. Accordingly, the foregoing description should be taken as illustrative and not in a limiting sense.

t' - M&C Folio No P51237GB 74

Claims

CLAIMS:

1. A user interface for controlling colour of colour input or output devices at multiple sites, which is implemented on a computer using object oriented components, comprising: an abstract menu object comprising: a command name object; means for drawing the command name object; means for selecting the abstract menu object; and a submenu command object derived from the abstract menu object, and comprising means for executing the submenu command object; a menu command object derived from the abstract menu object comprising: a plurality of the submenu command objects; means for selecting one of the submenu command objects; and a menu bar object derived from the abstract menu object comprising a plurality of the menu command objects, and a select menu for selecting one of the menu command objects; and a first action command object derived from the submenu command object comprising: means for selecting said action command objects; and means for carrying out said first action command object for defining color transformation of said color input or output devices in which at least one of said color devices is each located at at least two of said sites to enable each of said color devices at the sites to provide substantially the same color within the colors attainable by the device.

2. The user interface according to Claim I further comprising: a second action command derived from the submenu command object comprising: means for selecting said second action command object; and means for carrying out said second action command object for defining a network of said multiple sites.

I(,-, M&C Folio No P51237GB 75

3. The user interface according to Claim 2 wherein said means for carrying out said second action command object further comprises means for establishing a connection between said multiple sites for data communication through a shared file structure in which said shared file structure has one or more components capable of being shared by said sites.

4. The user interface according to Claim 2 wherein said means for carrying out said second action command object further comprises means for displaying the multiple sites of the network to the user as nodes linked to each other, and means for selecting connection between said sites.

5. The user interface according to Claim I further comprising: another action command object derived from the submenu command object comprising: means for selecting said another action command object; and means for carrying out said another action command object for defining user preferences for color reproduction by at least one of the color devices.

6. The user interface according to Claim 2 wherein said means for carrying out said another action command object further comprising means for enabling the user to select preferences of neutral color definition, %UCR, GCR, and maximum black.

7. The user interface according to Claim I fiirther comprising: another action command object derived from the submenu command object comprising: means for selecting said another action command object; and melins for carrying out said another action command object for displaying characteristics of said defined color transformation at one or more sites.

8. The user interface according to Claim 7 wherein said characteristics comprises at least the gamut of the color transformation of one or more color input or output devices.

9. The user interface according to Claim I further comprising: another action command derived from the submenu command object comprising: means for selecting said another action command object; and CP M&C Folio No P51237GB 76 means for carrying out said another action command object for comparing characteristics of said defined color transformation at one or more sites for two or more of said color devices,

10. The user interface according to Claim 9 wherein said characteristics comprises at least the gamut of the color transformation of one or more color input or output devices.

11. The user interface according to Claim I further comprising, another action command object derived from the submenu command object comprising: means for selecting said another action command object; and means for carrying out said another action command object for annotating images produced by at least one of said color devices.

12. The user interface according to Claim 2 wherein said network represents one of a telecommunications network and LAN.

13. The user interface according to Claim 2 wherein said network represents one of a Wide Area Network (WAN), and Internet.

14. The user interface according to Claim I wherein at least two of said sites are remote from each other.

15. A system for controlling color reproduction at multiple sites in which each of said sites has at teast one rendering device, said system comprising: a computer system having a user interface at at least one of said sites to enable a user at the site to select a plurality of sites and to connect said plurality of sites in a network; and means for enabling the production of information for transforming input color image data into output color image data for the rendering devices at at least two of said connected sites such that colors produced by the rendering devices appear substantially the same within output colors attainable by each of the rendering devices.

1 t - M&C Folio No P51237GB 77

16. A user interface for controlling color reproduction at multiple sites in which each of said sites has at least one rendering device and said user interface is operable at a computer comprising: means for enabling a user to select a plurality of sites and to connect said plurality of sites; and means for enabling the production of information for transfonning input color image data into output color image data for each of the rendering devices at each of said plurality of sites such that colors produced by the rendering devices appear substantially the same within output colors attainable by each of the rendering devices.