HDR Imaging

HDR imaging is a fairly new concept - it refers to High Dynamic Range.

This is fairly basic article - for a full, detailed and incredibly useful description of the process, replete with workflow suggestions for ordinary and panoramic image, please see Royce Howland's article at Naturescapes.net

What Is HDR?

HDR stands for High Dynamic Range and is a relatively new concept in photography. To understand what it is and why I'm so excited by it, here's a brief background.

The best, most expensive, camera you can buy today is but a toy compared to your eyes. Why? Because a camera (film or digital) can only record a very limited range of tones, whereas your eye can detect and transmit to your brain a vastly wider range of tones.

That's why you see a lovely landscape and take a quick picture of it and then get very disappointed with the result. If the foreground looks good, the sky is almost white. If the sky looks good then the ground is way too dark. If it's somewhere in between, then neither sky nor foreground looks that good.

This is simply because there is a limit to the dymanic range that film or a digital sensor can record. When presented with a scene with a wider range of tones, and most typical scenes have a wider range of tones) then something has to give. Depending on the exposure, the sky loses detail by being overexposed or the foreground loses detail by being underexposed.

Over the years, photographers have compensated for this by using grey graduated filters - filters where the top half is dark and the bottom half is clear. Using these, the sky can be darkened to the point where the dynamic range of the scene can now be recorded by the digital sensor/film.

This is low dynamic range photography - the grey grad filter lowers the dynamic range of the scene so the camera can capture it. It certainly looks better than a non-filtered image, but sometimes the filter's usage can be detected - for example, the sky being darker than its reflection in a lake, or the foreground clouds looking too dark.

Image Blending

With the advent of digital photography, programmers began to devise ways around the problem. In most imaging packages the Curves tool can be used to lighten the dark areas and darken the light areas. Providing the amound of adjustment isn't too large, this can yield good results.

The next stage up is image blending, where a number of images are taken at various exposures and are then blended together, taking the best bits from each image. Again, this can be quite successful, but it can fall down when there are too many differences in the series of images - the wind blowing trees/grass etc can prevent the blended images from being properly aligned.

A variation on this theme can be achieved using the RAW image format, which is a special format supported by most digital SLR's. The RAW format contains a greater dynamic range than the resulting 'developed' image and thus allows some post capture 'exposure compensation'. In particular, Canon EOS RAW images processed using Canon's own RAW development routines can often be 'developed' the +/- 2 stops exposure compensation. These can be blended with confidence that the alignment will be perfect. (This 'exposure compensation' is not as effective as taking a series of differently exposed images. It's much better than nothing, however.)

Blending techniques have their limitations - too aggressive and 'halos' appear where dark objects meet light opjects. To compensate for this, the blending routines have to smooth the transition between light and dark areas and this can lead to a loss of contrast and 'flat' images.

HDR

HDR appears to work in a similar fashion to blending, but with a subtle twist - rather than lowering the dynamic range to accomodate the limitations of film/the digtal sensor, it uses a range of images to determine the actual, higher, dynamic range of the scene.

In other words, when presented with a series of images of the same scene taken at, say, 2 stop intervals, the HDR process can determine (approximately) the real dynamic range difference between the land and the sky, between the darkest and lightest parts of the image.

The HDR format stores the resulting calculation as floating point numbers - this is an important distinction. A normal RGB image is stored as integer (whole) numbers. By using floating point numbers, the HDR image can much more accurately reflect the subtle differences between tones and this leads to better, more detailed final images.

The HDR image (currently) cannot be effectively displayed on screen. It needs to be processed into normal RGB integer data first. However, the nature of the HDR data means that the RGB output is more dynamic and effective than ordinary image blending, however sophisticated the blending process.

Is there a downside? Well, the best HDR images need to be planned - the photographer has to assess the scene and decide that it needs the HDR treatment and then capture a range of images of the same scene, typically at +2 stops, -2 stops and one in between these. Again, images with significant motion in them (the sea, moving branches, grass blowing in the wind) present difficulties, although the latest version of the HDR software I use is getting better at working with these.

I've also found that Canon RAW images, processed 'at capture', +2 and -2 exposure compensation provide very good results. Not quite as good as a pre-planned series of images, but far better than any blending technique I've used (or written). In such cases, it seems to help if the original RAW image is a little underexposed. The RAW exposure compensation then is able to provide both shadow and highlight detail in the +2 stops and -2 stops developed image. A grossly overexposed image would lack sufficient highlight information to be of much use.

The software I use for HDR is Photomatix. It comes as a standalone program or a Photoshop plugin and is worth every penny/cent. One tip - it can read most RAW formats, but I've found that processing the RAW image (as described above) yields better results. This is certainly the case with Canon RAW, where their RAW development routines produce very good 'exposure compensation' results. I can't speak for other RAW formats or Canon RAW processed using non-Canon software as I've not tested these.

An element of care does need to be excercised when using Photomatix - it has a large range of image adjustments available to it, and overdoing them can result in very unrealistic images. It's just a case of experimenting to find the settings that work with the images you are working with. Photomatix also seems to work best where the scene is quite 'contrasty' - images shot in flat light seem to benefit far less from it. In flat light, I still prefer to use an ND grad filter to bring the sky into the camera's dynamic range and a polarising filter can sometimes improve the color saturation (and skies) in these and other images.

A great feature of Photomatix is its batch processing - when processsing a set of similar images, the same Photomatix settings should work with all of them. Batch processing saves a stack of time.

Here's my Photomatix workflow with Canon RAW images, where I have NOT made a range of exposures:

1. In ARC (Andy's Raw Converter), set the white balance for the images
2. Set the HDR switch in ARC and send the images for RAW development. This automatically produces set of images at 0, +2 and -2 exposure compentation, suitably named for Photomatix to pick up in batch processing.
3. Open Photomatix and open the first (set of 3) images. It creates the HDR image.
4. In Photomatix, go to Tone Mapping and adjust the settings accordingly.
   1. If the images to be processed are similar I save the Photomatix settings and batch process the lot.
5. Once all the images have been HDR processed, I open the results in ARC and do whatever extra processing is needed. Usually, this is Curves and sharpening. Often I boost the saturation a little.
6. If I'm posting to the web, I will either use ARC web development feature to do the resize, framing and captioning at the same time as the post processing, or I use Image-Web-Ready to do this later.

Here's an example:

Original image - very underexposed, but the sky is OK. It's not easy to use a grey grad filter as it would interfere with the foreground too much:

After my HDR workflow: