My thoughts on HDR processing, and how to achieve the best results.
Firstly, I should say that this is not a ‘how to do HDR’ article. Here I assume you are already familiar with what HDR is and how to create a tone-mapped photograph. If not, there are many, many articles out there on the web to get you started.
HDR – the saviour of the poor photograph?
“I’ve just taken an OK image, so if I make an HDR out of it, it will look great – right?” Wrong. All the processing in the world will not make a great image out of a lousy one – composition is still everything, and the same ‘rules’ that apply to regular photography still apply to HDR. Equally, HDR is no substitute for great light, You can’t make beautiful light out of nothing, even with HDR!
So why use HDR at all? A good question, and some commentators would say – don’t – ever! Personally, I don’t subscribe to that point of view. Like all photographic techniques, HDR has its place, and finding novel ways to express something visually using HDR is part of the art of photography. I find that HDR lends itself particularly well to interior shots, where the enhanced microcontrast possible permits fantastic detail in brick and stone. It also permits the photographer to amplify the very subtle differences in light inside that are difficult to capture otherwise.
Photographers (should!) know that even the best cameras cannot capture the full range of tones visible to the naked eye, and that it is often necessary to apply some form of Dynamic Range Enhancement (DRE) to compensate for this. These techniques can range from – in the simplest case – the use of graduated filters to balance light across the frame or – in instances where the boundaries between very dark and very light areas are complex – underexposing then lightening shadows and mid-tones in one’s RAW editor. More complex that this are DRE techniques that capitalise on the use of multiple exposures to capture a wider range of tones; manual image blending being the simplest of these. HDR/tone mapping also fits into this category.
Turning up the volume
Even then, a photograph can only ever be an expression of reality rather than reality itself. For that reason, the photographer can only hope to capture an image in a way which ‘represents’ the way a scene really looks, in that way permitting viewers to connect with the photograph. And there are many ways of doing this, of which HDR is just one. One strength of HDR is in enhancing local contrast in very small areas which would otherwise look flat in a photography; ‘turning up the volume’ in the detail to literally ‘suck’ it out of a scene. But to do this well requires an understanding of how to ‘trick’ the viewer into believing a tone-mapped image is something real.
I use the industry standard Photomatix software for HDR creation and tone-mapping. Contrary to many HDR how to’s, I shoot bracketed sets of between 5 and 7 shots 1EV apart for combining into an HDR. I can’t say why this produces better results than the often quoted ‘3 at 2EV apart’, but clearly Photomatix has more information available within 7 frames. Whilst there will be far more overlap on tones, this extra information may help the software to better construct a more realistic and detailed HDR image. And please use a tripod for your bracketed sets! Having to rely on the software for aligning is only going degrade image quality.
Using Photomatix’s ‘Detail Enhacer’ algorithm, which is the default, the most important tone-mapping controls are Strength and Smoothing. The two are inter-related: up the strength and you need more smoothing to compensate. ‘Strength’ defines the degree to which the software algorithm will attempt to highlight detail across the tonal range. The use of strength is not rule-based, it is an artistic decision which applies to the image itself and how you wish to represent it. However, high strength requires more smoothing to prevent your image looking flat and ‘etched’.
Photomatix provides a histogram window during tone-mapping, and this is crucial to use for good results. The aim of the tone-mapping process is to squash detail from a wide dynamic range into a restricted range that can be displayed on screen and prints, and a clear, detailed image requires plenty of detail across this restricted tonal range. Ideally then, your tone curve should fill the entire histogram window, from dark to light, with plenty of height in the mid tones, although the exact shape of the curve will of course depend on the subject in question. As the aim of tone-mapping is to pull the maximum amount of detail from a bracketed set of images, if the tone curve only fills part of the available tonal space, clearly tonal detail is missing, and things are less than ideal.
Next in importance are the Black and White Point sliders. Experiment with these to expand the tone curve to fill the entire range. However, just like a regular photograph, if the tone curve is truncated at either end, you will have lost shadows or blown highlights. Think about tone-mapping as squashing a wide tone curve into a narrow one as fully as possible.
There are lots of other tone-mapping controls, but I have found them to be of little use. The next stage of processing in Photoshop provides far more control anyway. For example, don’t worry about colour saturation here. Photomatix tends to oversaturate anyway, so leave the slider midway, and adjust the saturation of individual colours in Photoshop.
One final thing. Don’t forget to save your tone-mapping settings as a starting point for next time. Use a name which reflects the subject. Remember that tone-mapping is more alchemy than science. There are no rules; the settings you choose will reflect how you wish to represent your subject matter.
So that’s me done then?
Erm – no! In fact you’ve just started, actually. Photomatix is very good at squashing detail from an extended tonal range into the restricted range required from viewing, but it’s lousy at making a good photograph. For this you need to post-process in Photoshop or your preferred image editing software.
Primarily, tone-mapped images are very soft. Sharpening is therefore essential. I mention this first as it is the step that will make the biggest difference to image quality, but it really should be left to the last stage in your workflow, as it also amplifies any noise. It’s worth noting that due to the softening caused by tone-mapping, lens quality has little to do with image quality in an HDR. The tone-mapping process will completely mask the inherent sharpness of your lens, so much so that it really doesn’t matter whether you are shooting with a kit lens or a top pro model. Any that’s good news for anyone shooting on a budget!
The big error in HDR processing
HDR enables even the darkest part of an image to be rendered brightly, highlighting detail in even the gloomiest places. But people often misinterpret this ability with efficacy, and remove all mood and emotion from a scene by making it too bright or flat. HDR images do not have to be bright! If you are taking an image of a dimly lit church, an over-brightened scene will just look artificial or contrived. The important bit…
All we are trying to do by making an HDR is capture detail across the full tonal range. Dark areas should still be dark, and dark scenes should still be dark – use just enough just light to make out a little detail in these areas.
Remember what you are trying to achieve. Your aim is to represent either the way a scene looked to you, or the way it made you feel, thereby evoking a similar emotion in the viewer. When I go on a location shoot, the first thing I don’t do is start taking photos. I’ll put the camera down and walk around a bit, trying to get a feel for the place, which will give me a memory from which to work when I come to process my images. Unless you remember a place, how can you possibly expect to make a good photo of it? And if you only see a place through a viewfinder, you’ll never get a feel for it.
Back to the tone curve
So it’s back to the tone curve again, to adjust shadows and mid-tones according to your requirements. I use Photoshop’s ‘Curves’ tool for this purpose. Again it is an art rather than a science. Just remember that as you experiment with adjusting the cures, keep an eye on the image, and continually ask yourself if this is the way it should look.
One technique I regularly use is to modify different regions of the image separately. It is often the case that e.g. a tone-curve adjustment to produce a dark moody sky would add too much contrast to the foreground subject. I will therefore create a selection of the sky, usually with the Quick Selection tool, adjust the tone curve here, then invert the selection to do the remainder of the image. I might even break the process down into even more steps, making multiple selections and doing them one by one.
A word of caution. Applying very different tone-curve adjustments to different parts of an image can cause artefacts along the boundaries between them. The feather of your selection edge is therefore critical. Too wide and the artefacts will be more visible, too narrow and the boundaries will look false. Unfortunately there are no hard and fast rules here. It really does depend on the subject in question, and experimentation is the key. Try between 3 and 7 pixels as a starting point.
And one final point here – it may be that some of your image does not need any curve adjustment, so keep that in mind, and don’t make adjustments for the sake of it.
As I mentioned previously, Photomatix tends to over-saturate colours, in fact it can often do a lousy job of colour rendition on the whole. Happily, Photoshop has some powerful tools for getting this right. I use the Hue/Saturation tool here, and will adjust the saturation (and sometimes also the tone) of each colour separately. Of course – if your monitor is not correctly calibrated, you may as well forget about this step – or do it with your eyes closed – for all the good it will do 😉
And now the fun bit. This is the point at which you really start to make something from your image. Using the Dodging and Burning techniques pioneered by Ansel Adams, and made desperately convenient in the digital darkroom, we begin to adjust local contrast according to our needs. For example, interior scenes will usually have internal corners, and if you look carefully at these, one wall will be lighter than the other. We can enhance this effect by darkening the shadows in the darker wall a bit with a soft brush, and lightening highlights in the other wall. This will emphasise the corner or edge, useful if we want to draw attention to it, e.g. to emphasise the converging lines in a corridor. A similar technique can also be used to highlight windows, in a similar manner to the way make-up can be used to draw attention to a person’s eyes.
I don’t often use layers for my adjustments, but this is one case where it really is good idea. It is very easy to make an image worse by dodging and burning, in much the same way that it’s easy to make a mess plastering a wall before you become adept. Having the altered image on a separate layer also permits you to switch the layer on and off repeatedly, thereby getting a good idea of the overall effect.
And finally its time to really see what your image looks like at its best, through sharpening. Photoshop has multiple sharpening tools, but like most people I prefer the Unsharp Mask tool. A great starting point for a high res image is a Radius of 1.5 and Amount 150. View your image at 100% when sharpening, and avoid over-sharpening to the point where the image looks etched. Finally, zoom in to see your entire image on screen, and switch your sharpened layer on and off a few times. You should still be able to see the difference. Only when you are sure should you flatten your image.
Of course, you are going to be keen to put your beautifully processed HDR photograph on one of your favourite photo-sharing websites, and for this you will create a JPG copy from your TIFF (you are working from a TIFF – aren’t you?!) then scale it down to a suitable screen resolution. And it’s here that people really ruin the look of their images.
Basically, reduced resolution images for the web always need resharpening. If you consider that sharpening is the enhancement of local contrast along detected edges, then resampling to a lower resolution will narrow the sharpened edges as well, ultimately rendering them invisible and making the image look fuzzy again.
Sharpening parameters should be matched to image resolution. For a web optimised image of width 1500px, I use the Smart Sharpen tool, Radius 0.7, Amount 90 as a starting point. Again, as above, review your sharpened image full size, and avoid over-sharpening.