medium, in order to enable improved interoperability within a single vendor’s products. Because PCS colorimetry may not be accurate relative to the original, the CMM cannot rely on the source colorimetry as represented in the PCS, and as a result v2 profiles will not support advanced CMM color rendering. There are also other issues that arise with v2 profiles because of the ambiguity of the v2 specification and incorrect interpretation of the specification in constructing profiles. 2.8 The ICC v4 Solution In ICC v4, colorimetric rendering intents are measurement based. They can therefore be relied on for proofing, and provide accurate colorimetry for CMM color re-rendering. Specification ambiguities are largely resolved and the text clarified to reduce the occurrence of incorrect implementations. A well-defined reference medium for the perceptual intent, with an asso- ciated gamut known as the Perceptual Reference Medium Gamut (PRMG), ensures cross- vendor interoperability. There is also greatly increased transform capability through extended look-up table (LUT) definitions, such as the lutAtoBtype which incorporates an additional matrix and curve and provides greater mathematical flexibility and an improved definition of 16-bit CIELAB. 2.9 ICC v4 Perceptual Intent Significant improvements have been made to the interoperability of the v4 perceptual path. The v4 perceptual intent color reproduction path is illustrated in Figure 2.3. With the PRMG, both input and output profiles can be based on a well-defined intermediate image colorimetry appropriate for the PCS reference medium and viewing conditions. The task of the CMM is thus to connect profiles with the same (or very similar) PCS gamuts, and minimal gamut mapping is required because the image colorimetry in the PCS is matched for the input and the output. Differences between source and output media color gamut and viewing condition are then dealt with consistently within the mapping to or from the reference medium performed by each profile. Perceptual intent color re-rendering Intermediate image colorimetry appropriate for PCS reference medium and viewing conditions Perceptual intent color re-rendering Input-side profile Output-side profile Source image Output image Figure 2.3 Perceptual intent color reproduction path in ICC v4 14 General The v4 perceptual transform includes both the data (typically device value) to PCS colorimetry transform, and color re-rendering to and from the reference medium in the PCS. The re-rendering operation includes consideration of: . differences in viewing conditions between source and reproduction and their appearance effects; . differences in media characteristics and image state; . color rendering preferences and the attributes of the preferred reproduction on the output medium. If the profiles incorporate all of these considerations, the task of the CMM is simply to connect the profiles together to create the transform between source and output data. The v4 perceptual t ransform is useful for general image reproduction across all devices and media. Since color re-rendering operations are typically proprietary, profiles from different sources may produce different “looks.” Users can then select profiles based on color re-re nderin g preferences. This was difficult before v4 due to the issues with the v2 specification described above and a lack of coordination between the different color manag ement components (the operating system, the applic ation, and the driver and/or output system raster image processor (RIP)). As differences between actual and reference media decrease, the perceptual and colorimetric intents should converge. Before v4, users were ca utious about the perceptual intent because of the inconsistencies with v2. However, it is still important that v4 profiles are correctly constructed and that color management is well coordinated in order to maximize the confidence of users. 2.10 ICC v4 Colorimetric Intents The ICC v4 colorimetric path is illustrated in Figure 2.4. The color gamut mapping performed by a v4 profile has three requirements: 1. The input data colorimetry should not be changed within the intersection of the input and output media gamuts. 2. Colors that are outside the source image gamut should not be produced in the o utput image. 3. Colors in the source image that are outside the output image gamut should be clipped. Colorimetric characterization Gamut mapping Colorimetric characterization Input-side profile Output-side profile Source image Output image Figure 2.4 ICC v4 colorimetric path Color Management – A Conceptual Overview 15 A colorimetric transform includes the device data to PCS colorimetry transform, based on measurements made using standard methods (as defined in ISO 13655 and described in Chapter 20 below). The transform also includes chromatic adaptation to and from the D50 PCS white point, if the data has a different reference white. This allows gamut mapping to be performed directly, if desired. In proofing situations, the extent of gamut mapping required is best minimized by the choice of proofing media. As the chromatic adaptation matrix is included in the profile, it is invertible if CMM-based chromatic adaptation is desired. The colorimetric intent does not include other appearance transforms, in order to avoid unnecessary color appearance model complexity, instability, and other issues mentioned above. Colorimetric transforms are useful for preview and proofing applications, and in support of CMM-based color rendering. The media-relative colorimetric with black point compensation (MRC þ BPC) provides a standard baseline CMM color rendering that is adequate when the media, substrate, and gamut shape differences are not large. This baseline reproduction model includes chromatic adaptation and media white relative colorimetry with black point scaling (on XYZ coordinates). It also includes gamut expansion and compression as required. The current widespread use of MRC þ BPC demonstrates the importance of media considerations. 2.11 ICC v4 CMM Color Rendering In ICC v4, it is possible for color rendering to be performed by the CMM rather than the profile, as illustrated in Figure 2.5. In this scenario, CMM algorithms col or re-render source image colorimetry to be appropriate for the actual output medium, takingintoconsideration source and output medium color gamuts and viewing conditions. They can also support color appearance model-based color re-rendering. CMM-based color rendering can take advantage of full output medium gamut, a nd facili tate user adju stment of color re-rendering at the time of output. For more details on CMM capabilities in ICC v4, see Chapters 6 and 31 below. Source RGB Source profile Working space profile Working space profile Display profile Working space profile CMYK destination profile CMYK destination profile Proofing system profile Working Space RGB Destination CMYK Proof (hard or soft copy) Editing preview (display RGB) Figure 2.5 ICC v4 CMM color rendering 16 General 2.12 Moving Forward Current research into color rendering supports both automated perceptual intent transform generation and selection, and increased CMM color rendering capability. High-quality ICC v4 tools and profiles need to become more widespread to move completely away from v2 issues. However, it is acknowledged that considerable work is still needed to fully coordinate color management across operating systems, applications, and devices. User interfaces also need considerable work, but should be based on v4 solutions rather than on codifying v2 problems, and shoul d ideally advance both color man agement and user interface effectiveness. ICC and its members and the color management community need to work in a coordinated way to advance all of the technologies described above, building where possible on solid understanding and communication. Clear and unambiguous definitions of color encoding and image state, for example through ISO 22028-1, are key elements of this process. Color Management – A Conceptual Overview 17 3 The Role of ICC Profiles in a Color Reproduction System 3.1 Introduction Color reproduction can be a complex process. There are many different color reproduction industries, often utilizing different media from one to another, and within some industries there may well be multiple media used. These different industries will often have differing reproduction requirements, even for the same image, depending on the reproduction process itself and the stage in the workflow at which the reproduction is made. For example, an image on a computer display may be required to accurately match the color of the original image, or be a pleasing (idealized) reproduction of that image, or be a color match to a printed reproduction of the original (soft-proofing), which in turn may be a color accurate or a pleasing rendition of the original. One of the most important decisions that has to be made by a user is what kind of reproduction is required at each stage of a workflow where a digital image is rendered in some form. In many systems where color reproduction is limited to a small number of input and output devices, the mathematical transformation which has to be applied to the image data to achieve the desired color is often heavily optimized for each device pair. In such situations the color reproduction requirements of each stage of the workflow are usually well defined, and the transformations are optimized for those requirements and the devices used. Although ICC profiles can be, and are, used to replicate such systems, many of them are based on proprietary algorithms – oft en utilizing measurement equipment specific to the system. In addition, the profiles may well provide procedures for fine tuning the algorithms. In the hands of reasonably skilled users these systems can – and do – produce results of very high quality. However, in workflows where multiple devices might be used, and particularly where the devices may not be known at the time of image capture or generation, proprietary systems are often impractical. It was primarily for such workflows that the specification for ICC profiles was established. Its goal is to provide a mechanism for defining the color of image data in a way that makes it possible to exchange images between systems, while retaining any color requirements imposed on the image. However, it needs to be recognized that ICC profiles Color Management: Understanding and Using ICC Profiles Edited by Phil Green Ó 2010 John Wiley & Sons, Ltd do not, by themselves, comprise a color reproduction system. An application that provides color management is required to utilize them – and each application may provide different levels of functionality in order to meet the particular requirements of the users in the market sector that the product is serving. So , as long as the user selects an imagi ng application appropriate to their needs, it should be possible to use ICC profiles to provide the desired color reproduction. Despite this, the ICC is sometimes criticized for various inadequacies of color reproduction. While some of the issues raised may be appropriate for attention by the ICC (and in most cases are being worked on by ICC Working Groups), others are more the responsibility of applications that use ICC profiles. In many cases, limitations and deficiencies encountered by users are those of the implemen tation, as opposed to the ICC specification. Some of the color reproduction issues are so dependent on the industry sector in which the images are being used that general solutions must be the responsibility of vendors and experts with experience in those markets. The ICC is a loose consortium of companies accommodating multiple industry sectors, and in many cases color reproduction solutions appropriate for one sector are not appropriate to others. Thus the ICC sees its main role as providing an open method to describe the color for each pixel of an image that needs to be matched, and a procedure for achieving that match. Where such a match is not desirable, the “best” solution is very difficult to define as it can depend on many factors. Thus, the best that the ICC can really offer are mechanisms (known as perceptual and saturation rendering intents) to enable a user to define that solution in a way that allows it to be communicated to others in the workflow, but not attempt to define how it is achieved. For this it is important that applications are selected that provide results suited to their needs. The correct use of the optimized perceptual and saturation renderings within each industry sector enables the production of high-quality reproductions, tailored to the user’s needs in that sector. These intents, together with the colorimetric renderings, enable many reproduction requirements to be met, and where an extension to the system is required for particular industry needs, vendors provide very sophisticated color reproduction systems. Such systems are based on the ICC specification, but include the additional tools demanded by the industry sectors they serve. Other vendors provide simpler systems that are easier to use, which serve other markets, often utilizing only the basic ICC architecture. Users need to verify that they are purchasing a system appropriate for their needs. 3.2 ICC Profiles – What Are They and How Are They Used? Each ICC input (or source) profile provides a number of color transformations (in the form of look-up tables, matrices, and/or curves) that define the color expected from the encoded data of the digital image, in an open format. In other words, the profile defines the color to be expected with any set of image values – which are often device values, but may be in some standard color image encoding (such as sRGB). The color space used by ICC profiles is the internationally accepted CIE system for defining color matches, so by using this it is possible to ensure that colors from input will match those on output (assuming the output has an adequate color gamut), for the viewing conditions for which the color is defined. The conditions selected by the ICC are those defined in international standards for viewing transparencies and prints; the resultant color space is known as the profile connection space (PCS). The fact that the format is 20 General public means that any ICC-compliant system should be able to use these profiles to interpret the color intended for that digital image. In conjunction with the correct display and/or output (destination) profiles, various reproduction options can be achieved. The reason why a profile contains multiple transforms is to allow the user to select the one appropriate for the purpose. The various rendering intents that these transforms provide are intended to be applicable to different reproduction goals. The choice can have a significant effect on the color reproduction achieved, so the selection of the appropriate transform is an important decision for the user. The basic way in which ICC profiles are typically used to achieve color reproduction is by combining a source profile with a dest ination profile to enable input data to be transformed to that required to give the required color on output. Selection of the appropriate transforms, by selection of the rendering intent, enables the desired reproduction to be achieved. The combining of the profiles is performed by a CMM, which can be provided at various places in the workflow (such as the image editing software, raster image processor, or printer driver, among others). In some reproduction procedures there may be more than two profiles used (such as simulating a print on a display), or even special cases where only one is used that has been constructed by combining a source and destination profile (DeviceLink profiles). However, these are natural extensions of the basic procedure described here and greater detail will be found in the ICC workflow guidelines. 3.3 ICC Profiles as Part of a Color Reproduction System Simply using a CMM that only supports the basic ICC architecture to calculate and apply the transformation from input device space to output device space does not necessarily provide a color reproduction system that suits all needs. So long as the application providing the CMM allows the selection of the appropriate renderi ng intents at the time when the appropriate profiles are combined, there are many market sectors where it is perfectly adequate – particularly where input devices are “smart.” However, there are other markets where it may not be. In such situations additional functionality needs to be provided by the color manage- ment vendor. 3.3.1 Image Editing One issue is that many captured images are not ideal. They frequently exhibit color casts, limited dynamic range, or poor tonal rendition, which may not be obvious on some media but will be when reproduced on others. Such “errors” need correcting during the process of reproduction. Algorithms for automatically optimizing digital images have been developed, and are a part of many image capture, color man agement, or editing applications. In fact they may often be applied without the user knowing. However, because of the subjective nature of color reproduction, such automatic algorithms may not suit every user, or every image. Thus, for high- quality imaging, unless the user is confident in the quality of captured images , every image should be assessed and corrected as necessary. Such corrections require a subjective assessment of the image, which means that it has to be rendered in some form to judge its quality. For many users a well-calibrated video display is adequate for this purpose, though for The Role of ICC Profiles in a Color Reproduction System 21 some high-quality applications the image is first rendered in its final form, which implies some sort of iterative correction process. Each ICC profile is defined for a specific combination of device and media (as appropriate) and as such, when used appropriately, should enable faithful reproduction of the colorimetry of the encoded image. Although the perceptual and saturation rendering intents include optimiza- tions for media and viewing condition differences, device profiles – which are determined independently of any images – do not apply image-specific optimizations. Where precision is of the utmost importance, color management software can be designed to update device profiles to also include image corrections, but because of the subjective nature of this correction it is usually sensible, in the view of many experts, to keep the characterization and image enhancement algorithms conceptually separate. Alternatively, the algorithms for image correction, if automated, can be applied at the same time as the media transform specified by the device profile. As “smart” CMMs (which add functionality by interpreting both profile and image information in calculating the reproduction transformation) are developed, such procedures are very likely. An input profile can be embedded in an image, or sent as a separate file. Either way it can be used to define the intended color as already stated. However, the sender of the file has to be responsible for ensuring that the correct profile is embedded, but equally importantly has the responsibility for ensuring that the image is pleasing. If the image needs correction this should be undertaken prior to sending it, by directly editing either the image or the profile. In the event that this has not been done, and it is the responsibility of the receiver to optimize the image to make it pleasing, this must be made clear when the image is sent. The sender of the file must then be prepared to accept the changes made, or ensure that a proofing cycle that will enable corrections to be specified is part of the workflow. 3.3.2 Rendering Issues The choice of rendering intent is an important one, as already discussed. General guidelines as to which rendering intent is appropriate to different types of images, and/or workflow stages, are given elsewhere in this book. Essentially the selection comes down to whether a colorimetric match is required between input and output, such as for proofing and preview applications (or when the output media have a gamut close to that of the image) or whether the reproduction is to be the most pleasing by compensating for the differences in viewing conditions and gamut between source and destina tion media. The different rendering intent transforms in a profile are usually dependent on the profile creation software used to make them. While colorimetric renderings may well be somewhat different – because different vendors can use different targets for profile creation, different measurement devices, and different mathematical models – such differences are usually small. However, the perceptual and saturation intents can vary significantly. With older profiles there was an additional complication concerning ambiguity around the definition of the white and black values in the PCS to which the appropriate image data should be mapped, which could be interpreted differently by different profiling vendors. Thus, when profiles from different vendors were combined, the results could be unpredictable and/o r low in quality. Although the use of Version 4 profiles shoul d avoid this latter issue, it is not intended to ensure that the perceptual and saturation rendering intent s provided by different vendors produce the same 22 General transformation. This is an area where different profiling vendors will provide solutions most appropriate to the markets they have most experience of, and it is up to the user to select that product which produces the most appropriate tables for their needs. The same vendor may even offer the option of different perceptual renderings to produce different “looks.” Differences between profiles will usually be mor e noticeable where the difference betwee n the source and destination gamut is large. To enable consistency of rendering on the input side, the ICC suggests the Perceptual Reference Medium Gamut as a rendering target for the perceptual rendering intent. If this is used in a rendering workflow, the output profile does not have to make arbitrary choices about how it maps the source gamut to the output medium gamut. One of the complications in trying to specify perceptual or colorimetric renderings in any objective way is the fact that there is limited agreement between experts as to what constitutes an optimum color re-rendering, which includes appearance and preference adjustments, and gamut mapping. Th is is complicated by the fact that such studies are inhe rently difficult. From the discussion above it will be clear that both media differences and image content affect the perceived quality of the color re-rendering, and separating these in any study is not easy. If both are included in the study it will generally be necessary to evaluate large numbers of images (maybe several hundred) before coming to a reasonable conclusion as to an optimum algorithm. The ICC sRGB v4 profile, for example, went through exhaustive testing by ICC members before it was adopted as a recommended solution to the perceptual intent transform from sRGB to the Perceptual Reference Medium Gamut. Even if we assume that the image has been edited to remove any problems – so that the profiles are only expected to optimize the mapping for the media differences – it is still difficult to get agreement on that mapping. Trying to find a single algorithm that will work well for a variety of source and destination media types, and for a range of gamut shapes, complicates that further. All these reasons, together with the fact that other issues (e.g., viewing condition differences and user preferences) are often compensated for in perceptual and saturation renderings, make it very difficult to come to any general recommendation on the way to perform such mappings. In general, users with high-quality expectations must choose their color management software with care, or rely on expertly designed systems provided by companies for specific markets. Such systems may well provide correction routines to enable users to achieve specific rendering of particular colors. 3.3.3 Retention of Separation Information One of the problems encountered in many practical color reproduction procedures is the difficulty of optimizing non-colorimetric profiles independe ntly of one another. Although this should be substantially eased by the use of v4 profiles, in which the PCS reference medium to be assumed in perceptual profiles is more precisely defined than previously, the wide differences in gamut and media which may be encountered between input and output, as well as the effect of image content, place a significant difficulty in the path of a vendor or user optimizing such profiles. While non-colorimetric renderings in profiles can be, and often are, optimized separately, the reproduction requirements of some high-quality market sectors require that final optimization can only be done for the pair of profiles to be employed in generating the color transformation. The Role of ICC Profiles in a Color Reproduction System 23 [...]... the updated 4.3 specification (ICC.1:2010) 4.1 Common Color Reproduction Goals Two common objectives in color reproduction are re-purposing and re-targeting Re-purposing is performed when color content that has been color rendered so that it is optimal for one output color encoding is subsequently color re-rendered to make it optimal for another output color encoding The typical display and printing of... follows: 1 Determine what type of colorimetry the input-side profile will be placing in the PCS (e.g., sRGB display colorimetry, photographic print colorimetry, photographic transparency colorimetry, oil painting colorimetry, etc.) This determination involves the nature of the original, and also any input-side color re-rendering that the profile may be performing Common Color Management Workflows and Rendering... effectively becomes the “original” which is to be matched in subsequent color management operations 4.2 Profile Functions ICC profiles (both v2 and v4) perform two functions The first, coordinate transformation, relates device color code values to colorimetric code values in the PCS The second, color rendering or color re-rendering, changes the colorimetry of an original to be better suited for some particular... measured colorimetry and apply a single method Also, some vendors did not provide true colorimetric transforms but incorporated an element of re-rendering into the colorimetric intents For example, black point scaling is common in v2 display profiles Because the v2 colorimetric intent PCS colorimetry cannot be relied on and it is not possible to get back to the original, unadapted source or destination colorimetry,... a Colorimetric Intent Image State tag [12] to the specification, which allows the image state to be specified for the colorimetric rendering intents When this tag is not present it is assumed that the colorimetric intent PCS colorimetry is “picture referred” – that is, it represents a hard copy or soft copy reproduction rather than original scene colorimetry The scene-referred signatures are scene colorimetry... colorimetry estimate, then the colorimetric intents can be used to convert the image to a scene-referred color encoding (or working space) If a colorimetric reproduction of the original scene were desired, then the ICC-absolute colorimetric rendering intent would be used for this conversion Alternatively, if some form of manual color rendering is to be performed, the media-relative colorimetric rendering intent... International Color Consortium, Reston, VA [10] ICC (2006) Profile Registry Proposal International Color Consortium, Reston, VA [11] ICC (2006) Colorimetric Intent Image State Tag Proposal International Color Consortium, Reston, VA [12] ICC (2006) New Technology Signatures Proposal International Color Consortium, Reston, VA [13] ICC (2006) Floating-Point Device Encoding Range Proposal International Color Consortium,... v2 sRGB profile Common Color Management Workflows and Rendering Intent Usage 31 A perceptual rendering, where the input-side transform color renders (e.g., a digital camera profile) or color re-renders (e.g., a “tuned” transparency scanner input profile) the colorimetry of a scene or an original to some proprietary virtual medium in the PCS (coordinate transform plus proprietary color re-rendering) These... transform The “absolute colorimetric” rendering intent is renamed “ICC-absolute colorimetric,” to avoid confusion with CIE terminology for colorimetry The ICC-absolute colorimetric rendering intent is calculated by the CMM from the media-relative colorimetric rendering intent in the profile, and is a coordinate transform that places CIE colorimetry (relative to a perfect reflecting diffuser illuminated by a D50... the “perceptual” and “saturation” rendering intents are intended for repurposing and the colorimetric rendering intents are intended for re-targeting operations such as proofing There are two specified colorimetric ICC rendering intents, media-relative colorimetric and ICC-absolute colorimetric, with the ICC-absolute colorimetric intended for cases in which the proof is desired to include the look of the . type of colorimetry the input-side profile will be placing in the PCS (e.g., sRGB display colorimetry, photographic print colorimetry, photographic transparency colorimetry, oil painting colorimetry,. image Output image Figure 2.4 ICC v4 colorimetric path Color Management – A Conceptual Overview 15 A colorimetric transform includes the device data to PCS colorimetry transform, based on measurements. (ICC.1:2010). 4.1 Common Color Reproduction Goals Two common objectives in color reproduction are re-purposing and re-targeting. Re-purposing is performed when color content that has been color rendered