Print, Pixels & Colour

Printers > Pixels & Colour

//mindmachine.co.uk/products/icons/MMA_Logo_s70p_off.pngNavigation Icons Guide
Printers - General IndexPrinters Index 
Printers General Methods IndexPrinter Methods Index
Brief item on Print Resolution

This section looks at principles common to many image handling technologies.

People generally get visual information from a computer via a screen or a printer. The two technologies have similarities.

On this Page
Pixels - Picture Elements
Pixels for Screens and Printers
Greyscale Halftoning and Dither
Dither Algortithms
Colour
Language Support
Colour Gamut

Pixels. Index - Overview

Almost all screens and printers build the image using picture elements or "pixels". 

Pixel pattern

A pixel is a point of light or ink sufficiently small to be unnoticed on its own. When an image is made from enough pixels the human eye tends to merge them and pick out the lines and spaces, rather than the individual dots. Pixels are normally arranged in a matrix - a rectangular grid. the pixels themselves might ideally be square but are often round. If a whole row of pixels are created by a scanning mechanism they might merge into a line.

A group of 7 x 5  pixels is just sufficient to convey basic upper-case "Latin" text if they are set dark on a printed page or light on a screen. Using a larger matrix of pixels allows lower case and more elaborate patterns. How good an image looks basically depends on how many pixels there are - more is better. Techniques like anti-aliasing and "ClearType" can help improve the look of images.

Pixelation can be used with both input and output devices. Images acquired by digital cameras and scanners are made up of pixels point-sampled from the material. Images reproduced by screens and printers reinterpret those point-samples as another set of pixels. There may be a one to one correspondence between the input and output pixels but usually the dots used differ in size.

Creating patterns from pixels suits both human sight and computer memory. For human sight as long as the individual pixels occupy a very small part of the visual field they seem to merge and normally pass unnoticed. Inside computer equipment each pixel is typically a bit, byte or word in memory, so they can easily be manipulated. By treating the whole of a screen or page as one big "bitmap" the computer can handle any sort of graphics.

The word "pixel" is derived from "picture element" and first became common in the 1960s; it seems to have been used intermittently before that time.

Index-Overview - Screen Resolutions
Computer screens on average have around a million pixels - typically either the 1024 x 768 XGA common on notebooks (786,432 pixels) or the 1280 x 1024 SXGA common on desktop screens (1,310,720 pixels). A screen typically holds something between 75 and 100 pixels per linear inch (3 or 4 per mm). There is a trend towards screens with higher esolutions of 300 pixels per inch for devices used close-too like mobile phones. There are also trends to larger screens with resolutions of 1600x1200 and to wide-screen formats like WUXGA with a 1920x1200  (2,304.000 pixels). Large screens are more difficult to make so technological progress makes them more readily available. 

Screen and print resolutions are often given in dots per inch (dpi). This possibly reflects the historic dominance of the US market but also perhaps a feeling that for this topic the millimetre is an inconveniently small unit whilst the metre is too large. If I write "12000 dots per metre" most people will do a doubletake before thinking "300dpi".

Screen sizes are sometimes given as a "resolution", without any size. For instance, a UXGA screen 1600 x 1200 - which might well mean it is physically large, though it could conceivably be palm sized with fine detail. More annoyingly screen sizes are often given in inches diagonal such as "22 inch" with no mention of the resolution. 


Index-Overview - Printer Resolution
Printers have much higher esolutions than screens. A page of quite ordinary looking A4 mono text commonly holds about 8 million pixels printed at 300dpi. (A4 is just over 8x11 inches but is normally printed with margins).

There are just a few printers with resolutions normally around 200 dpi - dot matrix and fax machines being examples. Both are known for giving crude print. The resolution normally accepted for text printing is 300 dpi. 

Dot Matrix Printers & PixelsPage on Dot Matrix Printers
Laser Printers & PixelsLaser Overview - Section on Pixels and Memory
Voxels - "Volume Elements".3D Printers and Voxels

Most recent printers are commonly capable of much higher figures. Laser printers range from 600 through to 2,400 dpi. Inkjets from 1200 to 9,600 dpi.

There are a few reasons for wanting smaller and more numerous pixels on paper.

Paper is used differently from a screen - people expect more information and hold the paper much closer to themselves.

Print technologies don't have a greyscale, so if colour or photographic material is wanted the computer makes up a dither pattern of pixels to give the appearance of one. More on dither patterns below

Screen resolution may not be good enough. Most people find eading on a screen quite tiring and there is evidence that the low resolution of screens lowers both understanding and recall of information.

There are some other potential uses for putting more pixels on a page - small print in legal contracts, convenient little engineering diagrams and suchlike. 

The main use for a pixel-count beyond 300 dpi is usually to provide greyscale by allowing the printer to use dither patterns.


Greyscale. Index-Overview - Greyscale

Pixels scanned from a view by a camera or from a page by a scanner have a property called greyscale; if they are monochrome they could be black or white or anything in between. This is dealt with by measuring and digitising the light level and communicating it using one or more bytes of data. If the image is colour then it will usually be sampled using red, green and blue filters or sub-pixels and each colour has it's own level. Colour information tends to need three times as much data.

This ability to handle different light levels is called "greyscale". The amount of light from a pixel on the screen can be scaled to show any shade of grey and not just black and white. On a colour screen each pixel divides into Red, Green and Blue sub-pixels which can be independently changed controlling colour-hue as well as the light level. About Colour  Most printers cannot do this.

Print processes are often described as "binary" - meaning a print dot or pixel has two states, either on or off. A dot matrix printer pin either fires or it does not. An inkjet bubble is either blown out or not. Laser printer toner is either present or not. Where colour touches the paper is saturated with it. If this seems peculiar remember that pens are like this - the pen either writes or it doesn't. It may be possible to make a fine line with a pen but the area written still gets a full coat of colour.

In the past, grey scale wasn't terribly important because text is black and white. Photography needs grey scale. Colour photography requires each of the contributing colour processes to be capable of grey scale.

Index-Overview - Greyscale and Halftoning
Half toning - Traditional print gets a grey scale by varying the dot-size.

The problem that ink can only transfer or not occurs in all kinds of printing, with the letterpress and offset lithe machines used for books and magazines for instance. Printed materials originally used various sorts of fine lines and cross-hatching to give the idea that an area should be grey. Later the idea of half toning developed. An image is reduced to patterns of small dots of ink of various sizes. The idea is that the dots are sufficiently small that the eye picks up the page area as shades of grey and doesn't notice individual dots. Making larger dots that cover more of the page gives darker shades.
 
Half toning for conventional printing was  done by optical methods in the plate etching process until the 1980s. The picture to be printed was imaged through a couple of glass screens which divided it into fine squares. The result is a  pattern made of dots of varying sizes.Haltone shading - general idea enlarged
The pattern is transferred to a photosensitive coat on the printing plate. The plate is etched and the halftone results in a fine pattern of bumps.

Problems that have to be addressed are that the eye may pick up the dot pattern in some circumstances - for instance a regularity in the original image might create a moire pattern. Changing the plate angle could improve matters. Halftoning for newsprint is often quite crude - at about 85 lines per inch. Magazines use halftone resolutions up to 185 lpi and coffee table art books push things further to 300lpi.

In the 1980s digital platemaking machines took over. A digital platemaker typically has a resolution of 2,400 dpi or more. This high resolution is mainly used to produce halftone patterns that won't distract the reader by producing moire patterns. The unaided human eye cannot make much of dots 100th of a millimetre across. 

Index-Overview - Greyscale -Dither
Dither - Digital Printers get a greyscale by changing the percentage of pixels printed.
 
Dither creates an effect similar to greyscale but more suited to computer printers. Halftone ink dots vary in size according to how dark the page is to be - merging into continual cover to make solid colour. Computer printers can't usually vary the size of pixels much but they can vary the percentage of pixels turned on and off. If the pixels are small enough that the eye merges them them this can give a similar effect. Turn half the pixels in an area black and the page looks neither black nor white but grey. Adjust the percentage of pixels black or white to get the desired effect.Checkerboard dither with half cells black and half white

Simple Dither Pattern

Dithering trades a devices spatial resolution for colouring ability. For instance, if a page of mono text looks good at 300 dpi then double the resolution to 600 dpi. Each pixel of the 300 dpi page is now 4 pixels at the higher resolution and by assigning ink or not to those positions there are five light levels - white through to all-dots-on being black

Dither isn't quite as simple as it sounds. Human sight is sensitive to something over a thousand shades of grey if they can be compared side by side. To reproduce a thousand levels of grey would imply using a 32 x 32 square of real but microscopic pixels clustered to represent one visually perceptible pixel (giving a cell with 1024 elements and 1025 intensities including all-white). Handling four colour CMYK print is even more difficult - with the problem varying depending on whether overprinting to produce product colours will work.

Printer designers could not simply multiply the number of pixels a thousand fold. Miniaturising the pixels isn't practical at all for dot matrix printers. Pins at about 0.1mm (1/250th inch) are already so fine they are vulnerable to breaking. Miniaturising the pixels produced by laser and inkjet printers is possible, but has been a matter of progressive improvement. In the early 1990s laser and inkjet printers were achieving the 300dpi resolution needed for plain text. Dividing each pixel into 32 x 32 subpixels might be something like an ideal, but printers with a 9000 dpi resolution do pose a few difficulties. Progress to 2,400 dpi on lasers and 5,000 dpi on inkjets has taken about 20 years of refinements to the design. Rather than use enormous resolution and a few full-intensity colour dots photographic printers tend to use light-cyan, light magenta and light -black (grey) inks.

Higher resolution print may be useful in other ways. For instance, it allows printing of very small text and finely detailed diagrams. Unless you are a lawyer working for a finance company or a very short sighted engineer the ability to print very small pixels is only really useful for dither patterns to make greyscale and colour.


Resolution and Greyscale limits. Index & Overview - Resolution & Greyscale Limits

One problem is that moving from 300 dpi to 600 dpi is not just a matter of halving the dimensions of the pixels. What is changing is an area, so it obeys a square law. Printers need more memory to hold the pixel patterns and more data communicated from the computer. In practice many printers with resolutions of 2,400 dpi or beyond actually generate the page from 1200 dpi data; anything more is unnecessary.

An inkjet head holds ink when it isn't firing and forms drops when it does fire by overcoming surface tension. Halve the diameter of the capillary and its sectional area quarters so the tension rises. In some ways this might be a good thing - random vibration is less likely to cause ink blots. However the energy needed to force ink out rises. So potentially  do the cavitation effects as bubbles collapse. The probability that the new narrower tube feeding ink will block with dry ink or grit also rises.

Laser printers hit different problems - collimation of the beam to a fine point focussed on the drum at different distances, faster spinning of the polygon mirror and closer spacing of charge elements. The net effect is similar, providing the high resolution to allow much higher esolution and a greyscale requires a process of improvement. Neither lasers or inkjets are yet capable of a completely convincing greyscale in all circumstances - but they may be the best technologies we now have, better than offset print and beginning to outpace silver halide photography.

Memory and Greyscale. Index - Overview - Memory - Greyscale

The amount of memory needed to support printer operation rises as a square. It takes a megabyte of memory to hold the 8 million pixels or so of a monochrome A4 sheet with a bit of margin at 300 dpi. Raise the resolution to 600dpi and it takes 4 megabytes of memory. Each doubling of resolution quadruples the memory, which in turn has to be handled by a processor and downloaded over network cables.

Improved printer resolutions are generally useful for giving finer lines but as resolutions go beyond 600dpi most human eyes can't pick out the detail being printed without a magnifying glass. Improvements in esolution mainly aim to provide the ability to use a dither pattern for greyscale.

Because pixels are being used represent greyscale print control software can assign one 24-bit colour perceptual pixel to a cell of maybe 64 eal pixels on the page. This cuts down the memory requirement, but at the cost that the resolution isn't available for small print and fine lines. One possibility is to have a print driver and print language capable of memory saving tricks like this but also of switching to handling fine lines in other areas of the page.


Dither Algorithm. Index - Overview - Dither Algorithm

Producing a dither pattern might be fairly simple, just a matter of eproducing the crosshatching or halftone pattern conventional printing would have used. Sight is rather sensitive to lines. It isn't sensible for the computer to fill small areas with lines or crosshatching in the old style because the viewers eye will pick this out and it will more likely be perceived as a pattern rather than grey. Dithers are often either very fine-pitched patterns or pseudo random in an attempt to defeat the eye's seeking for patterns.
 





As yet most printers aren't at the point where they can divide perceptible pixels at 300 dpi into dither patterns at a resolution of 300 x 32 = 9600 dpi and give a thousand-level greyscale. A simplistic approach wouldn't work all that well anyway. For instance:

At the nearly white and nearly black ends of the spectrum the appearance might be of random dots in the print rather than any perception of shade.

Pseudo random dither will tend to break up any lines and features in a picture - and since the eye is seeking these the result is to degrade the image.

Highly ordered bit patterns tend to create apparent lines and features that the picture should not have - and the eye picks these out.

Error diffusion dither is the approach generally used. An initial matrix is produced. Next the value (on / off) assigned to a pixel is measured against the required value and the error value distributed to surrounding pixels. The most common algorithm is a Floyd-Steinberg dither.

Printers algorithms try to make the number of dark pixels proportionate to the greyscale but also try to avoid visual artefacts - all this without equiring too large a cell.

Algorithms for producing dither might look at:

  • the actual range of shades present in the material,
  • at the presence and angle of contour lines,
  • at general rules for human perception
and if the image is colour - which it most likely is - at the general colour-space.

Dithering is one aspect of a general problem of quantisation - describing the apparently continuous nature of the real world properties using numbers or machinery limited to several discrete states.

Producing the dither will be the responsibility of the computer driver or printer RIP. Users with photographic requirements are often critical of the "graininess" of print, or of colour banding - particularly with inkjets. Both are likely products of the colour dithering mechanisms.
 
 
 

Index - Overview - Colour
Colour. - Dot MatrixDot Matrix ColourThermalThermal Printer ColourLaser Laser Printer Colour Inkjets Inkjets and Colour - Pigments and Dyes

Screen technologies use additive colour - the Red, Green and Blue colours mix and add up to white against the screen's more or less black background.

Printing inks subtract colours from white light. The Cyan, Magenta and Yellow should add up to black on the paper's more or less white background.

Actually the CMY printing inks often don't quite add up to a plain black - they make brown instead. Using three inks on one area might be rather more costly than need be. Black is commonly used for text covering most of a page so rather than pay for three inks to make one colour we add a fourth ink to give a solid black with just one ink. Most printers use black ink so the colour process used is CMYK.

Halftoning can also be used with a CMYK colour process. The colour dots have to be close to one another otherwise the eye won't associate them. The commercial print industry has various methods to produce the equired rosette patterns for the offset litho machines. A typical platemaking machine provides output at 2400 dpi or more to do this.

Dithering can also be used. Human sight is commonly said to be less sensitive to colour than it is to greyscale, distinguishing about a million colours and being rather more sensitive to greens. As it happens computers tend to work with 24 bit chunks and the upshot is that processes that need colour actually digitise red, green and blue to 8 bits each giving a 24 bit "truecolor" representation for cameras and screens. For printing this has to be converted to CMYK.

Colour print mechanisms can usually produce a range of 8 primary colours by combining the basic CMY range.
 
Colour 1Colour 2Colour 3Result
NoneNoneNoneWhite
CyanNoneNoneCyan
YellowNoneNoneYellow
MagentaNoneNoneMagenta
MagentaYellowNoneRed
CyanMagentaNoneBlue
CyanYellowNoneGreen
CyanMagentaYellowBlack

The 8 colour range isn't always useable - or there may be more than 8 colours. For instance with an inkjet overprinting magenta with yellow is likely to produce a ather different tone to overprinting yellow with magenta. Overprinting may also be unworkable because the colours don't mix, merely obscure one another, or because they produce haloes.
 
Even if the colours will combine the produced colours cannot include pink, orange or even grey. For those colours dithering is needed - Pink is pixels of magenta overprinting yellow mixed with pixels of plain white from the page.

Imperfections in the CMY process commonly produce a brown tinge rather than black but it isn't a guaranteed accurate colour.
Yellow / Cyan making green
The range of colours tends to be eferred to as a "gamut" or a colour-space. An 8 colour gamut brightens up diagrams but it won't do for photography.  The idea of a palette is related.This Yellow / Magenta square looks salmon pink on a monitor
To handle the million colours wanted by the eye most printers resort to dithering using their primary colour ange. Typically a printer will use a cell-size 5 x 5 or 8 x 8. If the native resolution of a printer is 1024 dpi then the colour resolution taking into account the 5 x 5 cells used for dithering might be a less obviously satisfactory 205 dpi. As for greyscale there are different algorithms for doing this. Magenta / Cyan should look Blue
Dithering done well should result in smooth transitions of colour like those seen in the sky looking smooth on the page. A badly done dither pattern will give distinct bands that weren't present in the original. Dither can also produce moire patterns - interference between the pattern of information and the dither pattern that create odd colours and "artefacts" on the page.Grey made from white balck 50/50 is likely to produce a moire effect

People looking at this page may well see moire patterns in the samples - interfering with their monitor of printer. LCD monitors are a bit prone to this with black and white squares like the sample above.

Printing True Greyscale.

Recent computer printers often can print greyscale to some extent, just not sufficiently well to be good enough on it's own. As photographic material has become commonplace it has become highly desirable to achieve some measure of greyscale.


Dye Sublimation. Details on Dye Sublimation Printers

The only common type of printer to manage a 256 level 8 bit greyscale matching 24 bit truecolor is the true "dye sublimation" type. Dye sub machines use a wax /resin colourant on a foil backing and this is exposed to powerful local heat. The original idea was that the colourant didn't just melt, it vaporised and condensed onto the paper - hence the name sublimation printer. What actually seems to happen is that the colourant may merely melt but its level of mobility varies with temperature.  By varying the high temperature used the printer can make a significant change in how much ink - colourant is transferred. This makes dye-sub machines great for photography.

Dye sublimation printers don't even need high resolution. The laser printer might use a 5 x 5 cell size for dithering, which gives it an effective esolution of just 205 dpi and more limited colour gamut. A dye-sub machine with a 300 dpi resolution can be more than a match for a 1200dpi colour laser. Laser printer and dye sub will both produce very good looking output, however, and only an expert is likely to be critical of either.

Dye sub can be quite cost effective for photographic material as well. Photographic material usually has near- complete page cover which the dye-sub foil can easily give.  For photographs dye sub could have an operating cost advantage over inkjet printers and may be cheaper to run than some laser printers.

Unfortunately the dye-sub printer is only really suited to pages which will be highly coloured and it just wastes any part of the foil that isn't used. This makes the dye-sub mechanism a rather costly way to produce text and a poor choice as sole printer for most users. The dye-sub printer is a useful addition to have around; only if the sole interest is photography can it be a substitute for inkjet or laser printing.

Greyscale in Other Printers

To simplify discussion the bulk of text above suggests that grey-scale is limited on most printers. This is largely true for users because the command set on most printers won't allow a grey except using either a transmitted or built in dither pattern.

Technically it is possible to get a dot-matrix to print a very limited grey scale.  The pin cannot be given a weak signal because it wouldn't impact the page, however at the end of a pass the printer can be told to make another identical pass which will transfer more ink. This technique of overprinting is quite commonly used to get bolder print- either built into the printer or by external software. Making two passes over the page does slow the printer down to half speed - and still doesn't give credible photographic quality. Dot Matrix and NLQ PrintDot Matrix and Colour

Laser printer OPCs do have some grey ability. The OPC material can vary it's electrical voltage if the light source is modulated. This has been used with text to provide anti-aliasing, where ounded corners are given a greyer pixel to remove the stepped effect that pixels can otherwise produce. A potential problem with using OPC greyscale ability is that it reduces the sharpness of the image but that may actually help produce gradations in pictures. Older photocopiers often have a button that varies the voltage settings to get some photo-reproduction ability at some cost to sharpness. Some but not all recent laser printers have between 120 and 256 level greyscales and these can have good colour reproduction. Laser Printers and GreyscaleDeveloper -Toner and analog responseGeneral item on lasers & greyscale  

Laser printers still struggle a bit with light tones, any mix of the straight cyan and magenta toners tends to give a rather dark colour.


Inkjets with thermal heads can vary the temperature and so can vary the dot-size, the results are similar to halftone print, with dots of varying size. Canon has apparently made a strong effort to get drop-size modulation from thermal heads. The problem seems to be one of linearity - ink that contacts the paper at all tends to immediately spread.

The inkjet print process has a big advantage over laser printers because it is relatively easy to add another ink into the mix, its just another cartridge and little other change has to be made to the mechanism. Inkjet designers can add light cyan and light magenta allowing a wider colour gamut and light tones to pictures.

Piezoelectric heads of some kinds are inherently capable of dot-size modulation. The piezoelectric effect is proportional to the voltage applied. This proportionality doesn't necessarily mean the ejected droplet is directly proportionate.  However varying the voltage does vary the dimension change, so wide analogue response should be possible. Another possibility might be to have several piezo elements in a chamber and use the combined effect, or a standing wave to eject material.


Language Support

Because most printer mechanisms only support binary print the print languages that control them originally  tended to do the same. Adobe PostScript Level 2 supports halftoning.


Colour Gamut.

Colour Gamut is the range of colours that a device can produce using whatever means it has. The term colour-space might be used instead - it means the same.

A simple way to explain colour gamut is in terms of the brightness of inks. If all the inks are dull then any combination of them can only produce dull colours. Even if pure ink is bright - fully transmitting its espective cyan, yellow or magenta colour - it may tend to lose some of that brightness as it sinks into the paper of the page.

Colour screens can only actually produce Red, Green and Blue sub-pixel colours but since most recent video adapters can control all three colours to 8 bit accuracy the screen can range through 16,777,216 colours. This would be more colours than the eye can detect - but it isn't quite enough for shades of grey so some devices can use 32 bit colour.

Colour printers typically have 4 ink colours - Cyan, Magenta, Yellow and Black. Including the white of the page and allowing the printer to mix colours this gives it a basic gamut of 8 colours. Given a 5 x 5 cell for dither patterns and 8 primary colours 25 x 8 might give a 400 colour gamut - although some patterns might be unusable.

To achieve a fuller range of colours the printer control software will need to define larger  cells within which it generates a dither pattern. The driver has to make a compromise between colour gamut and perceived esolution.

To improve the colour gamut on photo quality printers a manufacturer might incorporate light magenta and light cyan inks, giving a six colour printer. In theory light yellow might be useful, but it's perceived as a light colour anyway so it isn't much missed.
 
Some top end printers have Cyan, Magenta, Yellow, Light Cyan, Light Magenta, Black, Photographic Black and Grey. There are even models that add straight Red Green and Blue to this.


Summary.

Printing processes used by computers aren't generally capable of producing sufficient gradations of grey or  colour to satisfy human visual perception directly. Printers often either make a dot on the page or not.  Computer print processes do not easily produce the gradations in colour called grey-scale. Grey scale has become a desirable property so some laser and inkjet mechanisms can provide it - so can sublimation thermal heads.

By printing many small dots and relying on the human eye to combine them computer printers can produce any level of grey or any colour.  Conventional printing presses do this by using a half-tone process. The equivalent process using pseudo-random small dots is called dither.

Printing Cyan, Magenta and Yellow Dots that add up to a brown-black colour does work, but the toner costs a lot. Printers often add a black process that replaces that used for colour when it is too expensive and innacurate at producing darkness.

Using Light Cyan, Light Magenta and Grey colours gives a much brighter page in light coloured areas. The cost is that the printer needs more print mechanisms. Inkjet printers can fairly easily generate these extra colours. Laser printers have more expensive mechanisms and rarely use more than four colours.