THE ACCURATE REPRODUCTION OF SKIN TONES

Page 3

Adjusting the Image

The most serious weakness of RAW files is the paucity of colour measurement numbers available as you tweak the handles. Whereas in Photoshop itself you can have RGB, CMYK, Lab or HSB, in RAW you are stuck with RGB, the least useful of all! This forces the photographer away from finalising adjustments in RAW to using Adjustment Layers in Photoshop. We recommend Adjustment Layers rather than straight image adjustments so that you can blip the correction on and off to prevent colour fatigue and chromatic adaptation conning you. As you will see later it also opens up the possibility of selective colour corrections of both colour range and in different parts of the image.

Pink Babies and Red Grandads!

Now for the tricky bit. In spite of our best endeavours we still find under certain situations that flesh tones do not come through accurately, even under controlled conditions. Given that you have shot RAW, click balanced for the colour temperature of your lights and then adjusted your exposure in the RAW processor against your Macbeth Chart you might be forgiven for thinking all would be well. Fat chance! Most experienced digital photographers will tell you that if they shoot a wedding party, some people will be standing out like beacons in the crowd. Your editor should know he is one of them!

After some deliberation and searching through literally hundreds of medical research journals we are starting to get some pointers. There are two potential sources for the increase in redness. The colour of the skin is determined by the melanin concentration, its colour and also the proximity of the blood vessels carrying oxyhaemoglobin, the red oxygenated type. These blood vessels are buried under the outer layers of the skin and the penetration of light through this layer and back out again influences the colour appearance. The colour balance of the skin has been described as Tan versus Rosiness. Given that only red heads (aka your editor) have the red melanin (pheomelanin) it seems a reasonable hypothesis that they will turn up more red under stronger illumination. There is a further hypothesis which, to date we have not tested. Detectors in digital cameras are more sensitive to infrared light and to overcome this and prevent excess reds in the final image, a “hot” filter is often employed to shield the near infrared light (NIR). If this effect is balanced for the majority of scenes it seems reasonable that under some conditions, extra NIR might get back to the camera.

Regardless of the underlying reasons for the effect, the evidence is easy to collect. In the 4-up image shown, your editor is excessively red in images [1] and [2]. In image [3] the overall red balance has been adjusted using selective colour on the reds with a hue adjustment of +9 on the reds. However this has dragged the Macbeth light skin swatch too yellow – it is not responding to the camera in the same way at all. This is the classic selective red effect. To overcome it the mask of the Colour Adjustment Layer has been painted, to control where the colour adjustments are active [4].

Macbeth chart

Skin tone I

ABOVE: The graph shows the plotted data points for the girls above left. In the red ellipses the data from actual skin measurements are shown. The blue ellipses show where the calibrated image data lay. Note that Emily’s complexion is more yellow than that of Sally and that neither skin tone is rotated towards red from skin to image, even though there is a rotation towards red of the Macbeth swatch duen to the calibration profile. The estimated colour temperature was 5550°K. Overall the effects cause us to conclude that there is no over-reddening effect from either of the girl’s complexions.

In order to provide more background data we took the opportunity to photograph and measure the two young ladies featured in the picture above. We measured the skin specta of each girl and determined the Lab values on the forearms, fingers and the back of their hands. We are unsure of the accuracy of the technique but the two skin complexions were different in the way you would expect, with a rotation towards yellow for Emily (on the left). In the image file, which was corrected using a bespoke camera profile, Sally, on the right had hue angles of between 6° and 24° whilst Emily had angles between 24° and 27°. Both girls are relatively pale skinnned with saturation values of 15% for Sally and and 25% for Emily. This difference is just obvious in the original image and should survive through the magazine printing process.

And so to print

Having won your hard-earned, perfectly corrected image, the next hurdle is to get it out onto the print in one piece. In our experience, most profiled printers tend to reduce the saturation of the image while maintaining the hue – that is the colour shift is down a constant hue line. This is typified by the graph from the colour audit test target shown on the right. When a printer/paper cannot hold a high Dmax, the profile compensates by adjusting all saturations downwards. All of the above assumes that you honour the profile as you go to print. If you change up to a larger gamut volume the saturation will increase, if you change down, it will decrease; this is the cause of a lot of grief at external printing labs. Most unprofiled printers tend to reduce saturation; you may care to compensate for this before you send your file to print. If an unprofiled printer desaturates and is off colour balance you are likely to be in trouble, especially if the error is towards green or cyan. print responce to input ski tones

Baby Flesh Tones

We collected data from a study of a large number of babies. In general the values lie on the red side of average for Caucasians with saturations that do not rise above 35%. In Lab

A similar colour-shift effect to that described above has been observed when photographing very young children. Whilst we expect babies to be pink, many digital photographs show an excessive colour change on the forearms (and feet). It is a specific effect as both arms are affected even though they are catching different levels of light. The overall colour shift was too much for a commercial picture, forcing photographer, Martin Sellars, to selectively mask and modify the colour of the arms. As shown in the dual 3-D image, the corrections are quite subtle but make a huge difference to the way the picture looks. In Martin’s experience, the problem is specific to digital; no similar issues have arisen in decades of film shooting! Your Editor is presently working with the medical specialists to resolve the puzzle. The consultant dermatologist we questioned was not surprised and provided a cogent explanation of why babies have a higher venous blood flow nearer the surface of their little arms. However whilst babies’ arms are undoubtedly bluer, the scale of the effect is at odds with measurements on actual baby skin, as shown in the graph. It seems possible therefore that the greater accuracy of digital is emphasizing an effect, which has previously been masked by the properties of film. Sadly this leaves us with a solution requiring selective (time-consuming) masking. Martin points out that the majority of photographers mistakenly judge the colour to be too magenta which would require an increase in green to correct it. The bias is, in fact, blue and requires an increase in yellow to correct it. Martin adds about 10 points of yellow to the mid-tones and 5 points of yellow to the highlights in the Color Balance dialogue box. This is applied through a selection, feathered about 10 pixels, applied to the arms and using natural folds and shadows to hide the joins. He estimates about 10 seconds to effect the correction.

Correction

In the absence of any better advice as to the source of this problem we are currently unable to formulate a rapid solution. Discussions with quite a large number of photographers have shown that the effects described for babies, and the reddening of some adult complexions, are common. One photographer related a situation in which a model was no longer engaged for commissions because her skin always turned orange. Others mentioned the possibility that the UV-screen formulations, commonly built into modern cosmetics, were responsible for some abnormal effects. TV cameramen related that they have selectively filtered particular cameras or lights to improve skin rendering. What is evident is that we have stumbled into an area in which there is a lack of solid evidence. One thing we are sure of – the effect is not that which occurs when the wrong profile is assigned or attached to a digital image. This typically occurs if an sRGB file is moved up into a larger colour space such as Adobe RGB. This produces an overall increase in saturation which is not the effect we are studying, a rotation of the hue angle. Anything which increases saturation will make any hue shift more apparent, so the effects described will seem worse. It is the selective shifts that we are concerned with. We intend to keep hunting!

Conclusions

To some extent the pages leading up to this conclusion are simply legitimising the advice we are about to give. We hear so much guff about the effect of Adobe RGB and sRGB, this RAW file processor or that, this white balance setting versus another, etc etc that it is timely to put the record straight! Here is our advice summary:

Graph

• It matters little which colour space that you use as long as you honour the profile as you move the file around.

• Moving to a bigger colour space increases saturation, moving to a smaller colour space decreases saturation. In some instances going the “wrong” way actually makes a more pleasing final print!

• If you use RAW files you have a better chance of controlling the eventual result. Some JPEG algorithms do funny thing to your files and colour space – beware!

• If you work in Adobe RGB and fit your skin tones within the bounds shown in the graph on the third page of this feature you are reasonably certain to be in the right ballpark. Most problem files we are presented with are outside that range eg shooting with the incorrect white balance will shift you out of the range if you start near to a boundary. If you work with both Lab and HSB co-ordinates in the Info Palette you will get a better idea of what is going on with your skin tones.

• One sound piece of advice is to always ensure that your yellow value (measured in CMKY in the Info palette) is always higher than your magenta value and that your cyan lies at less than 1/3rd of the magenta value.

• The average values for white Caucasians are shown in the table for HSB and Lab and also for sRGB and Adobe RGB. Note that Gretag seem to be varying their notion of “average”, the new SG chart does not have the same values as the older Color Checker 24. Be aware that to most Caucasians a slight tan is preferred but

the opposite is true for most Asian, Oriental and South American cultures, who prize lighter skin and generally avoid exposure to the sun.

• Although babies are pinker than their, parents make sure that their skin tones do not drift towards blue magenta. • At the present time we have no answer to the red-head problem other than selective masking as described – we are working on it!

Chart 2

Page 1 - Page 2 - Page 3

What our members say
Why I like the Societies: SWPP gives my company a family to draw professional advice and support from. - Paul S
Find out more about the Societies here

Convention testimonials Thank you very much for letting me take part in this year’s SWPP Conference. I had a fantastic time and learnt much whilst assisting in the programme of events. Steve Broadley
Find out more about the Convention here

Photo Quote: A great photograph is one that fully expresses what one feels, in the deepest sense, about what is being photographed. - Ansel Adams

There are 227 days to get ready for the SWPP Convention and Trade Show at The Hilton London Metropole Hotel ...
which starts on Tuesday 8th January 2013