I dispositivi di input (scanner e fotocamera digitale) "vedono" qualunque colore, ma non è detto che lo vedano correttamente. Molti colori diversi possono generare la stessa risposta. Lo scanner di per sè non ha un gamut, ma ha un intervallo di risposta (response range).

Possiamo tuttavia creare un profilo per uno scanner perché, nonostante non abbia un gamut, il materiale di cui facciamo la scansione (carta o film) ne ha uno, o meglio l'intervallo di risposta dello scanner determina un gamut per il materiale.

Quindi si dovrà fare un profilo per ogni tipo di supporto dell'originale, quindi almeno uno per l'originale riflettente (carta fotografica), un altro per l'originale trasmissivo positivo (diapositiva) e un altro ancora per quello negativo. E per ognuno di questi originali ci possono essere versioni diverse (per esempio per i diversi tipi di diapositive). E si possono anche creare profili per materiale non fotografico (per esempio stampe offset, copertine di libri).

Le informazioni di base sui profili di input sono nella pagina Profilo ICC di input.

Per costruire il profilo dello scanner ci si deve procurare un target IT8 (generalmente fornito con i programmi di scansione e di profilazione) che può essere di tre tipi:

• 5" x 7" riflettente: una fotografia a colori 12.7 x 17.8 cm (IT8.7/2, ISO 12641);
• 4" x 5" trasparente: una pellicola negativa 10 x 12.7 cm a colori (IT8.7/1);
• 35 mm diapositiva: una diapositiva a colori (IT8.7/1).

In tutti i casi l'immagine riporta una serie di quadrati colorati, più eventuali altri soggetti. Questi quadrati descrivono il film o la carta fotografica a cui sono riferiti (non hanno nulla a che fare con lo scanner).

Qui sotto sono riportati i target riflettenti rispettivamente di Kodak, Agfa e Fuji. Ulteriori informazioni sul target IT si trovano qui.

The IT8 standards, and I am referring to the IT8.7/1 and IT8.7/2 transmissive and reflective targets, are for calibrating and profiling for scanning PHOTO FILM TRANSPARENCIES and PHOTO PRINTS. When used for this, they are very good. The grayscale at the bottom of the target represents the entire tonal range possible, FOR THAT MEDIA. The "white" patch (GS0) and the "black" patch (GS23) represent the minimum and maximum, respectively, densities possible in that medium. For that medium, there can be nothing lighter or darker.

The columns 13-15 are comprised of a 12 step scale of the pure cyan, magenta and yellow dyes, from the minimal perceivable density to the maximum density. Column 16 is a black scale made from the equal addition of the cyan, magenta and yellow dyes from columns 13-15. Columns 17-19 are the secondary colors, red, green and blue made from combining two of the dyes, in equal amounts, from columns 13-15. Thus, in columns 13-19, you have the pure scales made from the pure dyes, their secondaries, and their tertiary combinations.

The columns 1-12 are comprised of colors sampled from the L*C*h0 space at 12 hue angles, 3 lightness levels and 4 chroma steps. Columns 4, 8 and 12 are the maximum chromas possible at the selected hue and lightness levels. So columns 1-12 sample the ENTIRE GAMUT for the medium.

If you have followed this so far, and have some knowledge of color science, you can see that this is a very good representation of the color reproduction space for the medium of the target. It is possible to set your scanner for a variety of tonal reproductions and renderings, including a very exact perceptual rendering.

However, since each target is keyed to its medium, then individual targets are necessary for each medium (e.g. Kodak transparency, Kodak print, Fuji transparency, Fuji print, et cetera). For exacting reproduction, it would be necessary for Kodak, Fuji, Agfa and the other manufacturers to make targets for each material. Most have standardized on one transparency and one print medium.

Please remember to place the IT8 standards in their proper historical context. When the IT8 committee first drafted this standard, the ICC did not exist. The only color management available on a non-proprietary platform was ColorSync 1. That the IT8 standards have done so well for scanning calibration and profiling and lasted so long is a testament to a well done job by an unsung group of color experts. Instead of dumping on them and their work, we should thank them for their help.

The difficulty people are having with using the IT8 targets is not in scanning what they were designed for, but in scanning what they were not designed for. The gamuts possible with printed materials, cloth, paints, inks, etc. are different. Unfortunately, there are not any targets available for these media, with the exception of the new GretagMacbeth ColorChecker DC. This target is used for digital camera calibration and profiling. Due to its physical construction, it is not suitable for scanners since the colors do not lay flat to the scan surface and the target does not bend for attachment to a drum.

So, use the IT8 for what it was designed for, and understand that when its use is extended beyond its original intent, it may require some tweaking of the scanner profile to give satisfactory results.

Il profilo ICC di uno scanner viene creato con un programma apposito (alcuni sono elencati in fondo a questa pagina) assieme al quale viene generalmente fornito il target IT8 e un file con i dati di riferimento, cioè le coordinate Lab o XYZ di ogni quadrato colorato.

Il target IT8 viene letto con lo scanner a una risoluzione da 100 a 150 dpi se riflettente e di 1000 dpi se trasmittente. Vanno annullate tutte le possibili alterazioni del colore che il programma di scansione potrebbe fare: alcuni programmi di scansione (per esempio LinoColor) hanno un comando per eseguire una scansione e salvarla direttamente senza nessuna modifica del colore.

Il risultato della scansione viene memorizzato in un file (generalmente un TIFF) e viene analizzato dal programma che genera il profilo, il quale mette in corrispondenza i valori RGB letti dallo scanner con i valori Lab o XYZ di ogni tacca. Così è possibile prevedere le coordinate Lab o XYZ corrispondenti a una qualunque terna RGB.

Generalmente il profilo creato con un target IT8 è buono: l'errore rimane dentro  1 o 2 deltae. E le fluttuazioni dell'hardware di uno scanner influiscono poco sull'accuratezza e l'usabilità del profilo, per cui dovrebbe essere sufficiente verficarlo un paio di volte all'anno.

You have to realize there are at least two different ways a scanner can implement an ICC profile. If the software isn't ICC savvy, then yes, the option is to profile it and leave all the settings the same and pull the "raw" scan into Photoshop (hopefully in high bit or simply forget the profile). Then convert from scanner RGB to Working Space RGB using the profile.

A MUCH better way to work (if you wish) is to deal with a scanner that utilizes the input profile along with the display profile and allows you to edit high bit data IN the scanner software. LinoColor is a perfect example of this. You use the robust tools available in the software while the input (scanner profile) and display profile are providing a good soft proof. The Imacon FlexTight also allows this but far better then LinoColor (are you listening Heidelberg) is you can save out the high bit data after corrections with ColorFlex.

Yup, take a 14 bits per channel scanner like the Heidelberg Circon. This means a lot more shades per channel than you will need in the final 8 bit PostScript prepress transport format with the usual 16.7 million colors. So you can be sure to get the right shades per channel into the finescan - and with the intelligent ColorAssistant and concurrent prescanning and finescanning in JobAssistant things zip along just nicely. (Henrik Holmegaard)

In either case, you make decisions about what you want the image to look like. Method one has advantages in speeding up the scans but applying edits on high bit data means a file that is twice as large. You do get to use Photoshop's superb tool-set and interface. Method 2 allows you do get good data too, using the tools in the software and with better units, you can save out that corrected data in high bit as well. In either case, you have to make intelligent choices about how you want the file to "look."

ColorChecker DC (utilizzato come target per la creazione del profilo di una fotocamera digitale) non è un target adatto per la creazione del profilo di uno scanner per i tre buoni motivi che qui indica Robin Myers:

First, the chart is provided on a thin cardboard backing that almost guarantees that it will warp. I have yet to see a flat ColorChecker DC (or original ColorChecker for that matter). This can result in some patches being out of the plane of focus and result in color shifts (depending on the scanner). This has necessitated placing books on the top of the flatbed scanner cover to flatten the chart.

Second, the grid used to separate the color patches results in a mechanical separation of the patch surface from the plane of focus. As above, this can result in color shifts, depending on the scanner, its depth of field and the type of illumination.

Third, with the exception of the 8 glossy patches, all the remaining 232 patches have a matte surface. This is exactly what is needed for a camera target, to prevent reflections of the scene from the target surface back into the camera. However, matte finishes desaturate the color since the illumination is evenly reflected from the patch surface. Thus, the gamut of the ColorChecker DC may be less than that of the IT8 film targets (I will be verifying this very soon). A better design for a scanner target would consist of only glossy surface patches, thus increasing the target gamut and therefore providing a larger measure of the scanner's gamut. Remember though, the glossy surfaces are terrible for use with cameras because they act as small mirrors, reflecting the scene in front of the target back into the camera and ruining the calibration and profiling. (Robin Myers)

Tuttavia secondo C. David Tobie si può usare il target ColorChecker in uno scanner a letto piano (non in uno scanner a tamburo) per scopi speciali:

One of the reasons I acquired a ColorCheckerDC was to test not only digital camera profiling, but scanner profiling as well. It certainly works fine on a flatbed, either as a test file with known values to check extrapolation of IT8 based profiles, or as a special target for building scanner profiles. (C. David Tobie)

I scan the capture target at different gamma settings, and build a profile from each one. Then I assign the corresponding profile to each captured target and convert it to Lab. I compare these Lab values with those in the target description file, and choose the one that gives the lowest delta-e values, and use that gamma setting (or tone curve in a drum scanner) for all subsequent scans.(Bruce Fraser)

Interesting. Of course if the profiling system was perfect, you shouldn't see much difference between the results with different gamma settings, since it should be exactly compensating for any changed you make. Do you think in doing this you're finding the "sweet spot" of the profiling system, or are you optimizing the quantization of the RGB encoding used between the output of the scanner and the input of the CMM? (Graeme Gill)

Following this tangent of what has become a multi-headed hydra of a thread (Scum dots? We don't need no stinkin' scum dots!) relates to tone curves (or gamma settings) used to obtain optimal 8-bits per channel from a high-bit scanner prior to profiling. This is an an alternate strategy to Bruce Fraser's technique for finding the "sweet spot" of a scanner. It involves using curves or gammas such that the scanner produces RGB values for the IT8 grayscale that are linearly related to the L* values of the steps on the target. This spreads the RGB values out in a more uniformly perceptual way that may be more amenable to scanner profiling software. You can compute these "ideal" RGB target values like this (use a spreadsheet if you want):

For each IT8 grayscale step i (i = 1 through 22 across the bottom of the IT8 target):

RGBi = ((RGBmin - RGBmax)*(A16 - GSi))/(A16 - L16) + RGBmax

where:
RGBi is the target RGB value for grayscale step i
RGBmax is the RGB value you want for Dmin (e.g. 245)
RGBmin is the RGB value you want for Dmax (e.g. 5)
A16 is the L* value for patch A16 (from the IT8 reference file)
L16 is the L* value for patch L16 (from the IT8 reference file)
GSi is the L* value for grayscale step i (from the IT8 reference file)

So for example, looking at step 10 of the IT8 grayscale, then i = 10, and let's use RGBmax = 245, RGBmin = 5, A16 = 87, L16 = 7, GS10 = 51. The target RGB value for step 10 of the grayscale would then be:

RGB10 = ((5 - 245)*(87 - 51))/(87 - 7) + 245 = 137

If you repeat that calculation for *your* IT8 reference file values and *your* desired RGBmin and RGBmax, you get a table of target RGB values for all 22 IT8 grayscale steps. Then adjust your scanner tone curves or gammas so that these target RGB values are most closely achieved (making sure though, that you always get separation between adjacent grayscale steps!). Note that this technique is best used for a bit reduction tone curve (e.g. 16-bits down to 8-bits). Probably not such a good idea for 8-bit to 8-bit, since in that case you will lose tones.
Just an idea for the more adventurous types... (Bruce J. Lindbloom)

La Scanning Guide di Don Hutcheson si trova qui.

• ICC SCAN di ProfileCity (plug-in di Photoshop, $149)
• Heidelberg ScanOpen 4 (in bundle con LinoColor)
• GretagMacbeth ProfileMaker Pro Input (Mac e Win $500)
• ITEC ColorBlind Matchbox set di programmi per creare tutti i tipi di profili (Mac e Win $700)