Printin' talkin' blues. Or pinks and oranges

Colour. Dogs can’t see it. Humans can, even if Americans can’t spell it properly, but it’s hard to describe objectively because a lot of it is a sensation that’s processed inside the brain. Any child can name a huge range of colours, but describing and defining colour in non-subjective or non-comparative terms is surprisingly difficult.

Not convinced? Imagine your monochrome-visioned dog could talk and you want to describe your new red car to him, or indeed her.

“It looks like the way a tomato looks.”

“It doesn’t smell like a tomato.”

“No, red is the way light reflects off it. Like this red teeshirt.”

“But that smells of washing powder. And last night’s curry. And that nice woman we met in the park.”

“OK, that’s enough red.”

“She gave me a biscuit. Was that red?”

Just a fantasy, but it illustrates a real problem. Humans can talk about colour to other humans because we assume that everyone understands that tomatoes are red, bananas are yellow, clear skies are blue, grass is green and so on. But that doesn’t work when you try to communicate colour to something that uses different concepts – dogs pay more attention to their noses than their eyes, and computers work in numeric values for pixels and vector fills. Geoffrey Woolfe is working on a solution that will allow computers to understand non-technical human instructions for colour manipulation. He’s the principal scientist in the Xerox Innovation Group at Palo Alto, California.

“Today, especially in the office environment, there are many non-experts who know how they would like colour to appear but have no idea how to manipulate the colour to get what they want,” Woolfe says. “You shouldn’t have to be a colour expert to make the sky a deeper blue or add a bit of yellow to a sunset.” Humans can’t generally talk about colour numerically (apart from trained technicians who know that solid red has a value of 255,0,0 in RGB, 0,100,100,0 in CMYK or FF0000 in hexadecimal). If you use a measuring device such as a spectrophotometer it can show you numbers in its readout, but there’s no “natural” expression of colour in day-to-day language that a computer would understand. The nearest thing in our sector so far has been the Pantone reference sets and their rivals such as Munsell, but these only work for solid colours and tints, not instructions such as “make it more orangey in the mid-tones.”

That’s no problem for production staff working in graphic or web design or technicians in the print sector. They have learned how to define and manipulate colour numerically. The problem comes when “ordinary” people, usually the paying customers, need to communicate their requirements for colour. The goal of the Xerox colour scientists is to create a new human-computer interface language that will make adjusting colours in a document as easy as simply describing the colour. Users will be able to type “make the sky a deeper blue” or give a voice command “make the background carnation pink” and the software does the work. The invention, still in the research stage, creates “colour language” by translating human descriptions of colour into the precise numerical codes that machines use to print colour documents.

“I’m an imaging scientist and I’ve specialised in colour, and I was working on the nerdy end of colour, developing transforms and applications,” Woolfe explains. “When I came to Xerox I got put into a group that focused more on workflow, not so much the technical aspects of colour. That got me thinking, how do you work colour into a workflow where there are many players who touch a workflow, many of whom lack the technical expertise?” He’s described his answer to the problem in a paper called “Natural Language Colour Editing.” Xerox has filed for patents on this technology, even though it has yet to be turned into a completed language let alone a product. Woolfe says that colour adjustments could be made on devices like colour office printers and commercial presses without having to deal with the numbers.

Natural language means that common words or phrases could be typed or even spoken via a voice-recognition system. Something like cardinal red on a printer or monitor is expressed by a set of mathematical coordinates that identify a specific region in a three-dimensional space, which describes the complete the gamut of all the colours that the device can display or print. To make that colour less orange, a colour expert enters numbers that distort that original region to a new region in the gamut. Out in the non-technical world those concepts are meaningless. The ability to use common words to adjust colour would have important implications for non-experts as well as graphic artists, printers, photographers and anyone else who spends a significant amount of time fine-tuning colours in documents.

“Colour is something that everybody is very familiar with. Everybody uses words about colour and we have a great colour vocabulary,” Woolfe says. “But the applications that people write to manipulate colour are typically quite complex for ordinary users to understand. So my idea is to bridge that gap by translating the complex technology that we have into simple interfaces that everybody can understand, and words as a mechanism to do that.”

In his paper Woolfe explains that using colour names in natural language is the most widespread way of communicating colour. It uses common colour names, such as red, green, blue, and combinations of names to refine the specification, such as reddishbrown, greenish-blue, yellowish-green. There are also “modifying adjectives” to provide further subtle discrimination, such as light, dark, bright, saturated, vivid, muddy, moderate, dull, pale, or washed-out. So this is how people tend to talk when asking for colour modifications: They’ll say “make it slightly less yellow”, “much darker”, “more saturated”, “greener”, “punchier” and “a smidge lighter”.

So Woolfe focused on these common human descriptions of colour. He found common words used to distinguish different shades and colours could be mapped to the technical language of colour used in devices like colour office printers and commercial presses. “The innovative part of this is the mapping language,” Woolfe said. “At Xerox we’ve found that if you can connect the human dimension to the mathematical dimension, you get a lot of usability.”

One example would be colour proofing. “Often the customer will look at it and say ‘I don’t like it, I’d like to change the colour, then the printer goes back and does it again. This process goes back and forth and it costs a lot of money and a lot of time. Instead you could proof the document on a video system, send that back to the customer, have the customer compare the documents on video and then allow him to make the changes - he’s the person after all who knows best what he wants.”

The problem with doing that today is that the applications for editing the document are fairly complex, Woolfe points out. “A customer usually has to specify the changes to a pre-press or print professional who makes the changes and that’s where the costs and the time come in. If you can give the simple interface to the customer he can get what he wants, send it back to the printer and everybody is happy.”

So that’s the problem. Woolfe’s solution is to create tables that take these common phrases and interpret them for the colour range that they refer to within an image, and then the colour modification that they refer to. Once this has been done, a mask can be generated that defines the colours being referred to, and then applies a colour shift to just those areas. For example the phrase “make the blues slightly less purple” targets the blues in an image, with the word “slightly” meaning a small amount that’s given a pre-set numerical value, says 10%. So an automatic mask selects the blues, and a shift is applied that reduces red or magenta by 10% in these areas.

Woolfe starts with an existing dictionary of standardised colour names (NBS-ISCC), extended with some neutral grey colours, and then “maps” onto that the alternative words or names for similar colours that are used by the wider population. “In the end it’s all about usability,” Woolfe said. “Colour is so prevalent today, you shouldn’t have to be an expert to handle it.”

Once Woolfe has nailed down a natural language for colour, he says he will be able to apply the same principle to other things: “The chance for natural language to improve our interfaces extends well beyond colour. There are lots of things where people have got well developed vocabulary to express their intentions and their desires and we can translate that into the technology to change the way we print documents.”

So, you may never be able to discuss a sunset with your dog, but thanks to Woolfe you may soon be able to sing the blues to your inkjet.

Back to top