[3D Rendering Tutorial]
Understanding ScannersTom Arah warns that scanner manufacturers have a vested interest in pulling the wool over your eyes.
The first step involved in computer-based photographic work is more often than not to scan the image. It's not only the first step, it's also the most important. If the detail, range and accuracy aren't there in the scan it's always going to be an uphill struggle to produce good end results. In spite of this, very little thought is normally given to scanning. The process is treated as virtually transparent, rather like a file transfer but from paper to screen rather than from floppy to hard disk. I have to admit that this has tended to be my view - effectively that the work only really begins when the scan appears on screen. Recently however I decided to upgrade my scanner and to look at the issue in a bit more depth. The results have completely changed my mind.
The main reason that scanning is treated as a given is that the technology behind it is essentially straightforward. All digital scanners work on the same basic principle, shining light onto the artwork and then picking up the red, green and blue values of the reflected light on a CCD. It's completely different in the world of colour output where radically different printing technologies have led to ruthless competition and an evolution so rapid that the industry seems to reinvent itself from one year to the next. For colour scanning there have been no such obvious great leaps forward. Perhaps the only comparable watershed has been the development of single-pass scanning that means that the previously separate red, green and blue scans can be undertaken simultaneously.
Single-pass scanning is undoubtedly a big leap forward, but its major benefit is in terms of speed. This is welcome and important, but a scanner is never going to be used to the same extent as a printer so, for me at least, speed in itself is not reason enough to demand upgrading. In fact the only factor guaranteed to make me upgrade is a major improvement in scanning quality. Working on my general (and slightly embarrassing) rule that it's time to upgrade when the current standard is twice as good as the system you're currently using, I decided to investigate the three components of scanning quality - resolution, range and accuracy - to see if current scanners pass the test.
Looking at resolution first, it would initially seem to be an open and shut case. When scanners costing under a £100 advertise resolutions of 4800dpi and even 9600dpi, the 300dpi of my current Epson looks paltry. Of course though it's wise to be sceptical. The miniaturisation of elements on the CCD has certainly advanced over the years but is it realistic that they should be 32 times better? A resolution of 9600 dpi would enable a transparency to be blown up to over A3, but are these scanners really able to pick up that amount of detail?
The answer of course is no. CCDs do now pack in more elements per inch, but the true optical figures for desktop scanners are more likely to be 400 or 600 pixels per inch. The hugely inflated figures seen in adverts come from interpolation. The values for those additional pixels that are not optically and accurately determined are based on those pixels that are. Although these interpolated values are not randomly selected, it's important to realise that they are determined mathematically rather than optically. In fact the same basic process can be undertaken in software using Photoshop's resize image command and doing so shows that although results are reasonable at lower levels of interpolation, the more the image is blown up the more the system reveals its limitations.
The use of algorithm-based manipulation to offer more power is what computers are all about, but in the case of the interpolated resolution figures bandied about in adverts there's no doubt that a software feature is being used to conceal the limitations of the underlying hardware. In fact it's even worse than that. Resolution is important to quality but, apart from for 1-bit, black and white line-art images, it's not a one-to-one relationship. As we saw in last month's look at pre-press, the input resolution for photographic images should actually be determined by the halftone screen of the outputting device. As this seldom rises above 200lpi even when imagesetting the advantage of extra scanning resolution is purely that it allows the image to be scaled up without losing quality. While this is useful, unless you are doing a lot of poster work, it's not necessarily crucial.
To be honest, if the only advances made in scanning quality were those claimed for resolution I wouldn't be interested in upgrading. To begin with the improvement in resolution is nowhere near as big as claimed but more importantly it's not necessarily an improvement that will be useful. In other words, in spite of the advances, most photographic scanning will continue to be done in the basic 200 to 400dpi range. On the rare occasions when higher resolution is necessary I've always managed to cope in the past - either by blowing up line-art on a photocopier before scanning or by taking important images down to a service provider - and I would have reasonably happily continued to do so.
Ultimately though I did decide that the time had come to upgrade because of another advance you'll see a lot of in the advertisements - bit-depth. This is concerned with the tonal range of an image which ultimately comes down to the scanner's ability to discriminate between different colours. It's this tonal depth, rather than resolution, that is the really fundamental factor behind photographic image quality and so the claimed improvements are particularly important. In the past scanners were limited to 24-bits giving a total potential range of 16 million colours (2 to the power 24), but recent breakthroughs mean that scanners now offer 30-bit or even 36-bit processing giving potential ranges of 1,073,741,824 and 68,719,476,736 colours respectively. At 64 times or 4096 times better than the current standard this certainly passes my double the quality test!
With many scanners all apparently offering the same feature sets, it becomes particularly important to rely on reviews. In the group test in issue 24 the HP Scanjet 4C easily won the award for scanners under £1,000. By then it was already coming to the end of its product cycle so I decided to wait around for the next version. It was a long time coming but eventually the HP 6100C has arrived. Although the delay - and the numbering - suggested the skipping of a generation I was rather disappointed to find that the specification was essentially the same as for the 4C, namely 600dpi optical resolution and 30-bit processing. However based on another of my rules of thumb, that the latest release is unlikely to be worse than the previous, I decided to take the plunge.
First impressions of the scanner were overwhelmingly positive due mainly I must admit to the speed. Although I had thought that this would be relatively unimportant, to be able to scan a 2Mb grayscale image in less than 10 seconds really does make a huge difference. It not only stops scanning from being a chore it opens up whole new avenues such as the ability to experiment with different settings or to realistically scan correspondence into a business archive. Impressions of quality were also favourable. Detail in the shadows of a grayscale photograph were far superior to anything my existing Epson could achieve, while the onscreen colours of an image of a stained glass window were far brighter and cleaner than they were on paper.
But was the scan quality 48 times better than the original? The answer of course is no and would always have to be no for a very simple reason. After scanning and processing the image is still delivered to the photo-editing package as a 24-bit RGB file or an 8-bit grayscale file. In other words the final scanned image still has a maximum potential range of 16 million colours or 256 gray levels. So does this mean that bit-depth has no effect on quality and is completely meaningless?
In fact the increased bit-depth of the scanner's internal working does have an important role to play in terms of the tonal range of an image particularly when an image has to be adjusted. Essentially any time the brightness and contrast, exposure, emphasis or highlight and shadow tools are used, the scanner has to map original tones to modified tones. When this is done in a 24-bit colour space each time this happens different input tones will inevitably be mapped to the same output tones leading to gaps in the colour mapping. What this means in practice is that different values in the original will be clumped together resulting in eye-catching splotches where a solid area of a single colour is formed from what should be a range of tones.
This was the bane of my life with the Epson and I'm delighted to say that the problem is much better with the HP. However it's by no means perfect. Detail is still lost in very dark shadows so that solid black artefacts in particular appear in the scanned image that weren't there in the original. The only way that this effect can be avoided entirely is for the scanner to be able to optically discriminate between such similar tones. While the HP is better at this than the Epson, to pretend that it's able to discriminate 1 billion different hues is at best wishful thinking while for any desktop scanner to claim 16 billion should be a case for the courts. Ultimately 30-bit processing is very different from 30-bit discrimination and although it helps prevent increasing the inaccuracies inherent in tone mapping it does nothing to prevent them in the first place. Again software-based digital processing is being used to mask the underlying hardware limitations.
The last factor in scanning quality is colour accuracy and in many ways it's the most fundamental. After all there's little point talking about tonal depth and range if the tones are wrong in the first place. Ultimately scanning quality is completely built on colour accuracy. In a way this is a point so obvious that it hardly needs to be commented on - it's certainly an issue that the scanner manufacturers seem very happy to keep quiet. To take it for granted though is a big mistake. Earlier, for example, I talked about the favourable impression made by a scan looking brighter and more vivid than the original. On thinking about this a little more, I realised that in fact I should have been disappointed. The only thing a scanner should do is to give as accurate a reconstruction of the original as it can. Any tweaks to change and "improve" the image are best left to the user. Clearly the scanner had been set up by default to spread the range of colours too much. Rather like adding too much sugar or salt to your cooking it means that while the first taste has more impact the true flavours are getting lost.
To get the real flavours back I decided that I needed to investigate the scanner's colour matching further and to find out how I could go about recalibrating the scanner. I thought, for example, that it might be necessary to change the gamma of the green response curve to boost values at low levels while suppressing them at higher levels. To know exactly what was needed I would have to scan in known targets and compare their scanned values to those of the original. As the scanner didn't come with its own calibration system I had to improvise my own. I ended up scanning in pages from the hi-fi Pantone webcolor palette of internet-safe colours as this enabled me to compare the values of the scanned swatches as shown on Photoshop's Info palette against their official RGB formula.
I also investigated how I would go about changing the in-built response curves. There is some control offered over this, but the HP's Deskscan software seems determined to hide it away. In fact the program offers four in-built gamma settings - 1.8, 2.2, screen and colorsmart - but for some bizarre reason these are only accessible through the print paths dialog. More flexible control is available through adjusting the response curve directly in the Emphasis dialog it, but crucially changes cannot be made separately for the red, green and blue values. I thought I had finally found the control I would need when I found the colour adjustment command but, since it is never explained what adjusting the "distance" and "direction" on the dialog's coloured cube actually does, I can't say I'm sure!
In fact the real reason I never found out how the colour adjustment worked was that I never found out what I wanted to do with it. It was literally impossible to work out any pattern at all from my attempts at colour matching. On one page for example, the values of a scanned version of an orange (r255, g153, b0) were all uncannily accurate across all the gamma scans, while an orange brown (r204, g102, b0) three swatches away came through as a virtually unadulterated red! In other cases the green, red or blue values could all overshoot or undershoot by completely variable amounts. Ultimately the accuracy or otherwise of each red, green or blue value seemed to depend on a whole range of factors such as the other two levels and even factors like the paper stock. The one thing there wasn't was a clear relationship that could be built in to a colour correction curve. Hopefully in describing this I'm sounding reasonably calm, but at the time I was tearing my hair out in frustration.
Eventually it dawned on me that comparing the values of the scanned swatches to what they should be was the wrong way of looking at it. Instead I scanned in the pages with no gamma correction and compared the results to these values as well. This was a real eye-opener, in effect showing how the scanner actually saw the page before any correction took place. It wasn't a pretty sight but it explained a lot. For the orange (r255, g153, b0), for example, the no-gamma scanner values (r255, g80, b0) were around 50% off regarding the green but in all the gamma corrected scans this had been correctly upgraded. For the orange brown colour (r204, g102, b0), however, the no-gamma values (r150,g5,b0) were just too far off to be corrected. For the green in particular not enough information had been picked up in the first place to enable the necessary boosting.
Software v Hardware
By now this should be sounding familiar. What we have is the same old story of software-based digital manipulation being used to cover up hardware limitations. Being positive about this, there's no question that it's better than having no post-processing. The scans with no-gamma correction were by far the least accurate and I have to congratulate the HP software engineers on being able to massage them - for the most part - into such reasonable shape. On the other hand it's the third time that we've discovered digital processing being used to compensate for deficiencies in each of the core quality areas of resolution, tonal depth and colour accuracy. In each case manipulation can do a great deal, but it can never completely compensate for information that's just not there in the first place.
That's bad enough but I also think that it's not a coincidence that in each case the manufacturer benefits from the consumer's ignorance. After all the bottom line is that they want to sell kit. If we are naive enough to assume that a new 9600dpi scanner will be much more useful than a 600dpi one; or that 30-bit processing means that we'll suddenly have billions of colours to play with; or - most fundamental of all - that a new scanner will inevitably be more colour accurate because that's what scanning is all about, more fool us. In effect the scanning industry - or at least the drive to buy that fuels it - has been deliberately built on the customer's ignorance and naivety. This has not only been allowed to continue but has been actively encouraged.
Of course this isn't to say that the entire industry is a scam. Scanners now are certainly better than they were in the past. Overall I'd say that my new scanner does pass my double the quality test and at half the price of my old one. On the other hand it's certainly not the miracle that I'd been allowed and encouraged to believe. It'll be a long time before I next think about upgrading, but when I do I'll certainly be giving a lot more weight to the optical issues of scanning and I'll also be a great deal more sceptical.
Hopefully you've found the information you were looking for. For further information please click here.
For free trials and special offers please click the following recommended links:
For further information on the following design applications and subjects please click on the links below:
[3D], [3ds max], [Adobe], [Acrobat], [Cinema 4D], [Corel], [CorelDRAW], [Creative Suite], [Digital Image], [Dreamweaver], [Director], [Fireworks], [Flash], [FreeHand], [FrameMaker], [FrontPage], [GoLive], [Graphic Design], [HTML/CSS], [Illustrator], [InDesign], [Macromedia], [Macromedia Studio], [Microsoft], [NetObjects Fusion], [PageMaker], [Paint Shop Pro], [Painter], [Photo Editing], [PhotoImpact], [Photoshop], [Photoshop Elements], [Publisher], [QuarkXPress], [Web Design]
To continue your search on the designer-info.com site and beyond please use the Google and Amazon search boxes below:
|designer-info.com: independent, informed, intelligent, incisive, in-depth...|
All the work on the site (over 250 reviews, over 100 articles and tutorials) has been written by me, Tom Arah It's also me who maintains the site, answers your emails etc. The site is very popular and from your feedback I know it's a useful resource - but it takes a lot to keep it up.
You can help keep the site running, independent and free by Bookmarking the site (if you don't you might never find it again), telling others about it and by coming back (new content is added every month). Even better you can make a donation eg $5 the typical cost of just one issue of a print magazine or buy anything via Amazon.com or Amazon.co.uk (now or next time you feel like shopping) using these links or the designer-info.com shop - it's a great way of quickly finding the best buys, it costs you nothing and I gain a small but much-appreciated commission.
Thanks very much, Tom Arah
[DTP/Publishing] [Vector Drawing] [Bitmap/Photo] [Web] [3D]
[Articles/Tutorials] [Reviews/Archive] [Shop] [Home/What's New]