How to do a high-key portrait

Greetings photo buffs! I have had a little hiatus from my blog, but now I’m back with a little tutorial, namely ‘How to do a high-key portrait’.

This technique should be in every serious photographers bag of tricks, if it aint already, so here is the lay-down:

1) Use a good camera/ lens combination for portraiture. I used the full-frame Canon 5D Mark II and a Canon 85mm 1.8 at F/2.2.

2) Use soft, diffuse light. I shot these photos with natural light in-doors by a west-facing window in the middle of an overcast day in early April. Generally, you want very soft light for this effect, so use plenty large umbrellas/softboxes/diffusers, or north light, or an overcast day, or even shade. You do not want direct light or sun. Also, you want the background to be very bright – these were shot in front of a south-facing window.

3) Use low ISO. While you want to overexpose a tad (‘expose to the right’), use as low an ISO speed you can get away with for ultimate image quality. I shot at ISO 200. I’m applying the actual high-key effect later in Camera RAW, to avoid clipping the highlights in-camera. Your mileage may vary, though, so if you prefer, aim for the overexposed look directly, by dialing in ~ +2 stops of exposure compensation.

4) Use a large aperture. You want to get good seperation from face to background for that 3D effect, but you also want to nail sharpness on eyes. On a shot like this with the 85mm 1.8, f/2.2 was a good compromise.

5) Get at least one eye in focus. Rule of thumb. Eyes will draw the attention first, so make sure you focus on an eye. With portrait lenses and shallow Depth Of Field, often you won’t be able to hold both eyes in focus, particularly if the subject’s face is at an angle, and close to the camera. As long as one eye is in sharp focus, you’re OK. You should only break the sharp-eyes rule for artistic expression. (The first picture in this article actually breaks this rule, but I feel it gets away with it. Sofie was swaying a bit back and forth, so critical focus ended up on her eyebrow, while her eyes falls a bit out of focus, showing just how little DOF there is to work with).

6) Once shot, select your best frames and process expertly. Post processing is what makes a good picture a very good picture. I spared nothing in Camera Raw on these ones. First, I pushed the shots about 2 stops to overexpose them, but kept highlights in check. Then I turned down Saturation by a large amount, and increased the Vibrance by about the same. Those steps will give you this high-key, neutral, soft look. I did some skin spotting with the heal tool in Camera RAW. Finally, I added effects, such as Vignette, Split-toning and Film Grain to give the shots a ‘film-like’ appearance. I added a bit of skin diffusion (duplicate layer, large gaussian blur, Overlay blend mode, low opacity) and monochromatic contrast in Photoshop at the end.

And thats about all there is to it! The rest is up to you. Happy shooting ­čÖé

Photoshop: Magic Numbers?

Perhaps you already know what I am talking about. With the plethora of necessary dialogs and sliders in your daily Photoshop work, combined with a busy schedule and 600 photos to postprocess, how do you determine rational choices for each image, without batching everything with the same values?

From more than a decade of Photoshop work, I have noticed that I have settled on a few ‘magic numbers’ that at least eases my perfectionist mind as to what the ‘just right’ figure may be in several of Photoshops dialogs. Rationality is a very small part of these choices, they have more to do with making quick intuitive decisions that I won’t later regret.

The dialogs in question are in particular dialogs such as filters, for instance sharpening and blurring, the Fade dialog (which I use often), Layer Opacity for blending etc. and the magic numbers I always seem to settle on are these:

In percentage dialogs: 33% and 66%, in sliders, values of 90 and 127.

Why? This is magic, and magicians never devulge their secrets. Or, as we learn from The Hitchhikers Guide to the Galaxy, the answer to everything is… 42. No idea in arguing, just try it. The values of 90 and 127 are based on image dimensions (detail frequency) though, and are tailored to images in the ~6-15MP range.

Do you find that you always settle on the same figures in % and value dialogs? You don’t need to rationalize, but it would be very interesting to hear about your choices in a comment on this post. Don’t forget to include the general image size you are working with.

A new take on ‘available light’.

A friend, professional photographer Jann Lipka, has created this fine and concise video tutorial showing an example of how to use time and ‘available light’ to your advantage when creating product / studio shots.

“My Name is Jann Lipka, I’m a professional photographer from Stockholm, and I wanted to show you how to make use of your cell phone in a little bit different way…’

This video tutorial is also a part of this blog post on strobist which is about using lcd lightpanels of various kinds for illumination.

Here’s my first try, using my cellphone’s light, Canon40D, 30 second exposure, ISO 125:


Expanding the dynamic range of a single RAW file

This tutorial will focus on a technique tailored to expand the visible dynamic range of a single RAW file (as opposed to creating a HDR file or masking a bracketed sequence).

Most RAW editors has several tools to do just that built-in, but often you can not achieve your desired end result without complicated Photoshop work and masking techniques. In this tutorial, I’ll share some of my secrets for expanding the dynamic range of a single RAW file.

Observe the following photograph:

Single RAW file - developing for the shadows blow the highlights.

'Say cheese' Single RAW file - developing for the shadows blow the highlights.

This image is from a Canon 40D RAW file developed with Adobe’s Camera RAW 5.2. As the original image was underexposed, out of necessity to keep the highlights in range, I’ve added a positive exposure compensation to brighten the shadows, in turn leading to a clipping of the highlights. Even with tricks like Curves, Fill Light and Highlight Recovery, I cannot both expose the shadows as I want them, and also get great highlights – at least not without an unintentional hit in contrast in the process. My opinion is that neither Fill Light or Highlight Recovery are perfect in ACR, as by affecting the left or right third of the histogram respectively, they are both too broad, and don’t go far enough. – And they don’t give you precise control over the tonalities you effect.

Let’s look at my technique for dealing with this problem. For this tutorial we will be using ACR 5.2 as our RAW developer on a PC, as well as Photoshop CS4. The beauty of this technique is that it offers a more or less automated, lets call it ‘intuitive’, way of expanding your dynamic range without delving into Curves and complicated masking for every problem shot.

1 ) Develop your RAW image in ACR only for the shadows (left side of histogram). That is, concentrate on getting good shadow separation, and let the highlights blow for now (that means do not even try to recover the highlights, don’t touch the Recovery slider, better to let them blow out).

2 ) When satisfied, hold Shift key, and click ‘Open Object’ in ACR to open the image in Photoshop as a Smart Object.

3 ) In Photoshop, right-click the SO layer, and select ‘New Smart Object via Copy’.

4 ) Double-click the thumbnail of your new SO layer, to bring up the ACR dialog again.

5 ) This time, develop for the highlights (right side of histogram), and ignore the shadows (left side of histogram). Typically for this technique, there’ll be around 2 stops of difference between the two versions. Essentially this is just a matter of dragging the Exposure slider to the left.

6 ) Click ‘Done’ when satisfied that all highlights are recovered.

7 ) In Photoshop, now hide the top layer, then go to the Channels Palette and Ctrl-click the RGB channel thumbnail to select the highlights of your lower, lighter layer. What you’ve got now is a so-called Luminosity Mask.

8 ) Go back to the Layer Palette, turn on your top darker layer (select it if it isn’t selected), and click the ‘Add layer mask’ icon.

9 ) You should now see a composite of the two layers that will have expanded dynamic range, but also will look a bit lifeless and with poor contrast.

10 ) To add local contrast, you need to select your Layer Mask, and add Filter>Blur>Gaussian Blur. How much depends on a number of factors, including file size and detail frequency in your image. Typical key figures for Amount are in the 90-127 range. However, for some images, such as the sunset example in the end of this article, you may need to go as low as 1! This is up to your best judgement, and needs some experimentation and careful examination to set right for every image. You’ll want to achieve natural contrast and at the same time avoid visible halos and artifacts. Note that what you are doing here is essentially how Unsharp Mask worked in the dark rooms of yore. Only this digital equivalent is much, MUCH easier.

11 ) Finally, you may need to adjust the tonalities affected. You can do this by adding a Curves adjustment (Image>Adjustments>Curves) to the Layer Mask. Typically, you want to darken the mask a bit, by creating a point in the middle of the curve and dragging down about 2/5th of the way. You may also use other tools and adjustments on the mask to localize the effect – all non-destructively.

12 ) For a final kicker, now press Ctrl-Alt-Shift-e to create a new composite layer on top of the others, then add Filter>Sharpen>Unsharp Mask, Amount: 6-12, Radius: 90-127, Threshold: 0, to add a bit of extra local contrast. Typically I overdo this a little, then immediately go Edit>Fade to 66%. Alternatively, you can set the layer opacity of this composited contrast layer as you like.

End result:

The final result - highlights have now been smoothly recovered.

The final result - highlights have now been smoothly recovered.

Disclaimer: What you see above is no miracle cure for blown highlights. You will still need to expose your photo in a way so that all data can be recovered, and you might have to take care of increased shadow noise from pushing the shadow end.

Here’s another example comparing ACR’s Highlight Recovery to this method:

my method

Dynamic Range expansion - top: image developed for the shadows, middle: Recovery at maximum in ACR*, bottom: my method

* Recovery at 100, with positive exposure compensation of +2. A closer approximation of ‘my method’ could be achieved by maxing out both Recovery and Fill Light in ACR, leaving Exposure slider at 0, but doing so introduces nasty edge artifacts and very low contrast.


I’ve compiled this technique into a Photoshop action, which you may download and study here.

Copyright 2008 Mathias Vejerslev

A Virtual Tripod with Photoshop CS4 Extended

This techique describes a method to stack several high ISO source images in Photoshop CS3 or CS4 Extended, in order to minimize random noise – in effect creating a synthesized long duration low-ISO exposure.

It is the first in a series of techniques relating to ‘computational photography’ – the process of utilizing several source images and raw computer power to enhance your final image output.

So you are at an art museum. You’ve got your shiny new DSLR with you, but the strict rules of the museum excludes both tripod and flash use. Unfortunately the art pieces are not all that well lit, maybe just the equivalent of a single incandescent lightbulb from the distance of a couple of meters.

– But you really want to get that depiction of a museum piece… What do you do?

My Canon 40D features a maximum of 6.5 frames per second. Instead of taking one exposure at high ISO, I can take a sequence of exposures at high ISO, and then combine all of them in a Photoshop Stack to create an image devoid of the rampant random noise. Photoshop aligns, then averages the pixels from each shot to create noise free pixels, and extract details otherwise obscured. The reason this works is because the noise is random – it is rarely in the same spot in two seperate images.

Astronomers has stacked images like this for a long time, indeed they have some very sophisticated software available to them, but now everyone with a copy of Photoshop Extended can utilize its powerful align and stack algorithms to do this trick.

Here’s the workflow:

1) Set an exposure suitable for the lighting and for handholding. For a painting, you don’t need much DOF, so an open aperture is suitable, just choose the first aperture that is sharp for your lens. In this example, my camera settings were: RAW, 1/160s f 2.2 ISO 3200. I want a fast shutterspeed for my 50mm 1.4 to ensure sharpness across the handheld sequence I’m about to make. I choose the highest ISO setting on my camera, ignoring any noise issues for now.

2) Angle the shot, focus, brace the camera as best you can, and hold down the shutter for as long as necessary. The more shots you have, the better a result you can expect, but you should also expect a longer processing time. My Canon 40D filled its buffer at 15 photos at 10MP RAW format, and it took it just a bit over 2 seconds to do so.

3) So, now you’ve got 15 noisy, virtually useless, slightly underexposed ISO 3200 shots of the same washed out painting. Great.

At this point, back at my computer, I open up Bridge, and examine details with the Loupe on all of the frames. If one or two are out of focus, I delete them or exclude them. Then I open up all of the remaining sharp shots in Camera Raw. I make sure to select them all, dial in a +1 exposure, because I had to underexpose the shots a bit to get a faster shutter speed. This will push my source images to ISO 6400(!). I carefully adjust the WB and other settings. I like to keep a linear tone curve for this technique, but your mileage may vary. You may even add a bit of capture sharpening, disregarding the hit you take in noise. – But do turn off all noise reduction, as we will deal with this in the next step.

I usually keep to a 16-bit Prophoto workflow, but I find that for a 15 image stack in Photoshop, it might just be to much data crunching to wait for, so I made my source images 8-bit Adobe RGB in this case. See screenshot:

Camera RAW - 15 photos selected, note positive exposure compensation.

Camera RAW - 15 photos selected, note positive exposure compensation.

Note that these are not photos of an actual piece of art, but of a 1 euro reproduction print I bought in a tourist trap in Spain. It’s hanging on a wall in my room, and it came in handy for illustrating this article.

Next, I click Done.

4) In Photoshop, go File>Scripts>Load Files into Stack. Select your images, and make sure to check ‘Attempt to automatically Align Source Images’ and ‘Create Smart Object after loading layers’.

If you have a lot of source images, this process can take a little while. Go make a cup of coffee, pet your cat or sumthin’.

5) When done, go Layer>Smart Objects>Stack Mode>Mean.

The pixels of all the layers will be mean blended, and the noise will be seriously diminished. Taking into account my photos were ISO 6400 originally, they now look great. I now flatten my image (Layer>Flatten Image), clean up any residual noise with Neat Image, apply a transform to straighten up the image, then apply a healthy dose of Filter>Sharpen>Smart Sharpen. The final step is a slight crop, because when handholding for a sequence like this, framing will differ somewhat, and around the edges of the photo there’ll be a margin where there’s more noise, because there are fewer overlapping frames.

Here’s some before / after crops at 100%:

Comparison of 1 original ISO 6400 frame vs stacked, Neat Image + Smart Sharpen result.

Comparison of 1 original ISO 6400 frame vs stacked, Neat Image + Smart Sharpen result.

And here is the final result:

Result of 15 image mean stack

Result of 15 image mean stack

It sounds like a lot of trouble, but once you establish this workflow, it is quite easy. And it can get you those shots that would otherwise be impossible, such as in museums or other low light situations with static subjects, where you prefer not to use, or cant use, a flash or a tripod.

If you think this sounds like to much work, here’s an idea for a $1 camera stabilizer.