Not quite HDR
Though it pains me, I have to admit that I got HDR wrong. I was even quite cocky about it three (?) years ago when I began using what was then the very new technique of tonemapping High Dynamic Range images. Enough to want to write a book about it, as I saw this as the answer to a fundamental problem faced by all photographers doing interiors. The problem: most interior views have too high a dynamic range (we used to call it contrast range then) for even film to capture. The old answer: bring in lighting.
A colonial interior shot on film, with bounced lighting behind and to one side of the camera to fill in some of the shadows.
Indeed, as film gave way to digital, the problem got worse, because highlights, such as through a daylit window, instead of smoothly fading and flaring, got clipped, with a sharp, aggressive edge that shouted ‘look at me’. But at the same time, the good thing about digital was that we could shoot a bracketed range of exposures in total register, to the pixel, and that paved the way for combining a range of exposures into one (admittedly unviewable) image file. From there, it was inevitable that people would work on ways to combine all this information, from extreme highlights to deep shadows.
Anyway, this is not a treatise on HDR, but rather a look at ways of capturing a wider dynamic range than a camera sensor is capable of — and then delivering it as a photograph that looks like, well, a photograph. The principle is simple, and probably most people reading this know it already. You keep the camera as steady as possible, ideally on a tripod, shoot a sequence of exposures about 2 ƒ-stops apart so that the darkest captures all highlight detail and the lightest captures the deepest shadows as mid-tones. If nothing else, at this stage you have archived the entire dynamic range of the scene.
An exposure sequence with an interval of 2 f-stops. The darkest captures all highlight detail, the lightest captures deep shadows as mid-tones.
Obviously this works best, you might even say only, for static scenes. In any case, this is classic interiors territory, and for the uses to which most interior photography is put, potentially very valuable. These are interiors magazines, ditto coffee-table books, and commercial uses such as hotel brochures and architects’ portfolios, all of which usually demand maximum clarity. Enter HDR, even though all the early work on it was done for something else entirely different, namely high-end computer graphics for the motion picture industry. Actually, making a single HDR-format file from a sequence of frames is not at all complicated. What is difficult is making that image file viewable on a normal monitor or in a print, and for this the answer is called tonemapping. There are many solutions, called algorithms for deciding what value each pixel is going to be when this 32-bit-per-channel file is compressed, as it must be, and none of them look completely realistic. But after the academics had initiated some early algorithms, people like Adobe and Photomatix worked on the tonemapping, gradually making improvements.
I got it wrong by blithely assuming that everyone would, like me, see HDR as an answer to a problem. Or at least, the minority of photographers for whom it was a problem. Not at all, as it turned out. All the things that seemed to me to be teething troubles, including haloing, plasticky textures and a generally hyper-ventilated look that was definitely not photographic, were happily seized upon by thousands of photographers on Flickr and elsewhere. Large numbers of people actually wanted – still want – pictures to look like this…. http://www.flickr.com/photos/13kingdoms/255352551/ and http://www.flickr.com/groups/hdr/ where there are, yes, 52,000 members. And who am I to say it looks tacky? It just wasn’t what I was thinking about when I wrote this book Mastering HDR Photography.
It wasn’t particularly what the software developers were thinking about either, and I remain in correspondence with Geraldine Joffre at Photomatix, who must of course deal with both sides of the market. She reminded me that there are no objective criteria for whether an image looks false or natural, and that if blown-out highlights and unresolvable shadows look natural to some (many?) people, it is because that is what we have come to expect from a photograph in the past.
So for photographers who want to maintain the traditional look of a photograph but still work with a higher captured dynamic range, what are the options? HDR remains one, but with a great deal of work, and almost inevitably some blending of different strengths and algorithms. The most popular software remains Photomatix, although there are others. Here, for example, is one I did for a forthcoming architectural book.
The Oyster Bar in New York’s Grand Central Station. A sequence of 4 exposures converted into a single HDR image, then tonemapped in Photomatix.
Now, the software is constantly being improved, but I have to say that there came a point at which I simply grew tired of the effort needed to keep the image looking natural. Geraldine and I had discussions about the meaning of natural and realistic. While there are many views following different personal tastes, it seems to have polarised across photographers into those for whom the ideal is looking like a photograph, and those who want images to look the way they think they see. There’s a tricky point here, because the Human Vision System does not see all detail in a view all at once. It builds up, very rapidly, a composite view from different parts, and this includes brightness extremes. So really, there is no possibility of producing a flat image that reproduces exactly how we see the original scene. I subscribe to the first group – I want photographs to look like photographs.
Fortunately for me, Photomatix then produced an alternative way of combining a sequence of differently exposed frames, calling it Exposure Blending and later changing the name to Exposure Fusion. Now this also has a number of sliders that need experimentation, though less complex than tonemapping. It also cannot retain all the detail that tonemapping can from extremes of brightness. But it has one quality, which is that the blended image retains a photographic look. I don’t particularly like recommending software, but in this case I make an exception.
A sequence of 4 exposures...
.
.... Blended using Photomatix Exposure Fusion.
Of course, given that my needs, and those of most professionals, are quite practical and not experimental, even more basic ways of blending frames can produce the results. One of these is simply to make a stack of layers and erase by brush out the bits you don’t want. This is much more akin to dodging and burning than software algorithms, because you do it by eye and by hand. It can be as straightforward a matter as taking just two frames, one exposed for the shadows, the other for the highlights. If the bright highlights and the deep shadows are in separate areas of the image, and if there are no exact hard edges containing them, selective erasing is simple.
The same 4 frames in a layer stack in Photoshop...
.
... Selectively erased by hand with a brush.
.
The final manually blended result.
Finally, all this is premised on wanting to hold all detail everywhere. Professionally, as in interiors shots, this is often the case, but creatively not necessarily so. Shadow areas can retain their mystery – look at the high-contrast printing of Bill Brandt, for example, or some of the images by Laura El-Tantawy, our featured photographer who appears after Tim Rudman.
