Color Profiling, an Idiots Guide to Getting the Right Color

By David J Foster

I have a scanner, a monitor, an inkjet printer and a printing press...I have a picture that comes in via the scanner, is previewed on my monitor, proofed on my inkjet and sent to press for final printing. Wouldn't it be nice if the picture looked the same at each stage of the process?

The honest answer is that it won't - but you can minimize the differences.

The scanner and the monitor are RGB devices - the monitor creates color by ADDING red, green and blue light. The more light they produce, the brighter or more saturated the color. In theory there is no limit to the color that you can produce this way, imagine a screen made up with lasers! The monitor can therefore produce a great range of colors (gamut).

The printers are CMYK devices, they take the whiteness of the substrate and SUBTRACT colour by applying colored filters. RGB have a bigger gamut than CMYK.

We have to make RGB colors look like CMYK.

The trick is to use a 'device-independent' color space, this is a theoretical model where all colours are ascribed a position in a 3D matrix, the axes are L, a & b. Now we can take an colour on the picture, the scanner will reference a database of colors and find an Lab value that corresponds to that colour, because it has a database it can also give an RGB value but that RGB value is specific to the device.

The monitor will look at the Lab value and produce a different RGB value, but crucially, those colours should look the same. In the same way the CMYK devices look up the Lab value and translate it into a CMYK value.

However, we said that RGB have a larger color gamut than CMYK, that is they can print a greater range of colors. What do we do when one of these out-of-gamut colors has to be printed on a CMYK printer. Well it depends what you want to do with it.

I might have an advertising/POS job with a picture of a lady wearing a bright pink with some text and a company logo. The company would like their logo to be their corporate colour so although I can't get the exact colour I should be as close as possible, to this end I will move the out-of-gamut colour to the nearest available colour. This will satisfy the logo, but that pink dress which is mainly out-of-gamut, if I move all the colours to the nearest point I may find that a lot of those pinks actually end up at the same colour and the dress will look like one flat colour. Clearly here the absolute colour is not critical but the differences between the colours are crucial. That's to say I'd rather print a dress of the wrong colour that showed the subtleties of shade between the folds than a flat colour which was more accurate but looks unrealistic. Instead of moving the out-of-gamut colours to the nearest in-gamut colour, we would take the whole range of colours and concertina them into the available gamut. Not only are the out-of-gamut colours moved but every colour, even those that are in gamut, gets shifted along a fraction. Thus even corporate logos that we can reproduce accurately are deliberately changed.

These are our rendering intents - when you value colour fidelity over shading use relative or absolute colorimetric (there's a minor difference between them, one is relative to the paper, one isn't).

When correct colour is of secondary importance use perceptual rendering.

With perceptual rendering it is obviously an undesirable thing to deliberately change a corporate colour and with colorimetric rendering we can end up ruining our dresses (in a manner of speaking). If you have a decent RIP, and I recommend Wasatch SoftRIP, RGB and CMYK portions of the file are handled separately. Send the pictures as RGB & process them with perceptual rendering, send the corporate stuff as CMYK & use colorimetric.

It won't make everything right, but it will make it less wrong. And to be less wrong is an ambition too...

David Foster has 22 years experience in the pre-press industry and specialises in sales & support of imagesetters for the UK screenprint market.