More on Processing
Last month I looked at alternatives to caffeinated HDR processing as ways of handling a wide dynamic range — and a bracketed sequence of exposures. Since then, I’ve had occasion to put these into practice as my assistant and I prepare the images for my new book on the Tea-Horse Road, shot largely in southwestern China and Tibet, and I realised that there’s still more to say. Processing is, in any case, an open-ended subject, and it’s this ‘endless’ aspect of it — you can go on tinkering and re-interpreting raw images forever — that has put me off writing a book about it! Nevertheless, as we all know, it makes a difference.
From the days of shooting Kodachrome, for which only a handful of labs around the world were prepared or able to push or pull the processing by spectacular one-third of a stop, to the latest generation of raw converter software, the role of processing has changed fundamentally. Because we can now completely alter the character of a shot during processing, there’s an obligation to get properly involved in it. Like many obligations, I’m not completely happy about it, because at times, particularly the end of a trip or the prep stage for a new book, it threatens to take over my day. Admittedly not as bad as missing daylight completely by spending it in a wet darkroom, but every shot that you think is important now carries with it a processing penalty. You can click Auto somewhere on the screen and be done with it, but you know that it can be better than that.
This is how it goes in my office. Let’s say we’re at the stage of having selected all the worthwhile frames from a take, and are ready to process. The raw images are archived and all are in one catalogue (we use Expression Media). One by one, we review them, enlarged on the screen. What we’re actually looking at, of course, is a quick conversion made by the cataloguing software, so at this stage we’re not pretending to look at subtleties. Without entering any discussion about actual detailed processing — brushes, vibrance, clarity and so on — we talk about how the image should look. By and large this is a non-technical discussion. I say what I was thinking at the time of shooting, or what I was hoping to achieve, and we look at it in the cold light of the computer room. I might say, for instance, that I think we should keep some shadows deep and strong so that the brighter subject comes forward more. Yukako might say that the shading from light to dark in one area ought to be kept subtle, as she knows that I have a tendency to go for stronger contrast than she does.
A first look at a Raw image in an image database.
We might have a long conversation about one particular hue. The range between bright orange and yellow often causes me problems — meaning I want to veer away from too much orange. For each camera, we’ve constructed and tweaked different profiles, and these tend to take care of this particular issue, but often we’ll compare the results from profiles when we get to open the raw image in the raw converter.
Basically, my answer to the too-full range of processing possibilities is to step back from the software and just think about how the image ought to look, irrespective of how this could be achieved. Between Yukako and me, we know technically how to produce any effect, from subtle to strong, so what becomes more important is a question of style.
This came up recently with one image in particular, shot in the summer, in the Tibetan part of western Sichuan, on the high grasslands (around 4,000 metres). It was the end of the day, and the light ultra-clear. At this altitude the air is noticeably thin, so that a low sun cuts through in a way completely unfamiliar to us lowlanders. And in this instance, there was a skyscape of the kind that I have rarely seen — a towering cloud system in rapid motion, with huge contrast and an almost unbelievable three-dimensionality. Better still, there was a landscape of some interest underneath it.
The range, however, was way beyond the sensor — at a guess, 14 ƒstops from the brightest white in the clouds to the deepest shadow on the hillside. I shot a bracketed sequence, starting with an exposure dark enough to avoid any highlight clipping, then increasing the exposure frame by frame in 2-stop increments until the deepest shadows were recorded at mid-tone. This is the usual shooting range for an HDR sequence, although HDR and tonemapping is not what I planned. I simply wanted to reproduce the sensation I had at the time, that afternoon.
The exposure sequence, 2 ƒstops apart. The central frame, outlined in red, was chosen for single-frame processing, as below.
This kind if sequence is as much for insurance as anything else, because with careful single-frame processing, a lot can be pulled out. Taking the central exposure from the strip above, the Raw image opens like this in default in Photoshop…
One Raw file opened in ACR and given the Photoshop Default treatment.
…but we can do much better at holding the range by using global controls like Recovery and Fill Light, and even more by using either the Adjustment Brush or Graduated Filter to perform the digital equivalent of dodging and burning…
The single image given the optimal treatment.
Now this avoids clipping, but tools like Recovery and Fill Light are digital workarounds to give the appearance of a larger dynamic range, and in fact quite a lot of detail is lost in the cloud highlights and hill shadows. There is a certain flatness in both areas because of the exposure compromise and these digital recovery procedures. A richer result can be created by using all of the five original exposures. The basis for my final result, below, is Photomatix’ Exposure Fusion, as I described last month. And it is just the basis, because I later combine this with some better-held areas from two of the originals…
My final choice of processing, using all of the exposures, and taking about 45 minutes.
In particular, the clouds have more three-dimensionality, and the grassland comes alive. But what could HDR tonemapping have done? The possibilities are endless, given the number of sliders and different algorithms available, but let’s quickly look at two defaults offered each by Photomatix and by Photoshop…
The default settings in Photomatix ‘Details Enhancer’ tonemapping from an HDR file. This is local tonemapping. A veil-like appearance overall, which could be dealt with later, and every tone held, but more painterly than photographic.
.
The default settings in Photomatix ‘Tone Compressor’ tonemapping from an HDR file. This is global tonemapping, has a more ‘photo-realistic’ appearance but much less effective in holding the dynamic range. The ground of the landscape, however, looks good.
.
Photoshop’s tonemapping option Equalize Histogram. Another global procedure, like the one immediately above.
.
Photoshop’s Local Adaptation tonemapping method, with the default settings. Much more could be done by adjusting sliders and the histogram, but the results generally trade efficiency for less ‘photo-realism’, as is typical with all these local tonemapping methods.
Tonemapping has become more sophisticated and controllable over the last two or three years, and each of the above has something that could be a starting point, although none of them — for my purposes — on its own. Obviously, they are each hugely adjustable.
But back to the processing workflow in general, or rather my processing workflow. Even though my assistant and I do this preview together, I start much earlier. On a trip (the one I just returned from was six weeks), I like to work on a few selected images on my laptop, just to see how they will look, and also to do it while the idea is still fresh in my mind. I don’t ever expect the results to be perfect, because the variable viewing conditions and the screen don’t match the 30-inch 14-bit Eizo that we use in London, but I can get reasonably close — even here, on a small boat on the upper Irrawaddy (yes, a little extreme, but there was plenty of time on that trip).
On-the-spot processing on a laptop, not pretending accuracy, but while the scene is fresh in the mind.
I do this on the smallest size of outputted image that the raw converter allows, and this is particularly valuable when it comes to local corrections. ACR in Photoshop is my personal choice of software, and several applications of the Adjustment Brush can add up to quite a complex processing log. This log, in the form of the sidecar file, is a valuable timesaver, and I always make sure that this gets saved in the backups. Later, in London, opening these once-processed raw files at hi-res and on the big screen, gives me the processing choices that I took earlier when I was travelling. If nothing else, these are a valuable reminder of what I was thinking then.
And there, in fact, lies the important component of processing for me— what I was thinking at the time. Whichever processing software you use, you’ll have noticed that it has been acquiring more capabilities and more options as time goes by and new versions come up. This is the nature of software: constant improvement, or if not exactly that, then at least tinkering. The now large range of processing treatment that you can give to an image throws the ball back to something more basic, namely how you think an image ought to appear. That, as I see it, means a choice between four quite different aims. If you think about each of these when it comes to processing any image, you may find the results will be surprisingly varied:-
- As it measures objectively (if you have the chance to do this, using a ColorChecker and detailed exposure readings, normally only in a studio)
- As you remember seeing it, objectively (which because of the way we see would mean equalising the shadows and highlights in some way — the main argument for processing methods such as HDR)
- With a ‘photographic’ appearance (that is, with the tonal and colour feeling that you would expect from film).
- Idealised (grass greener, skies bluer, that sort of thing).
