A thing or two about Dynamic Range…

This is not an another post about what is Dynamic Range, introduction to HDR and other such things. There are several articles on the Internets for that.

This article is about the usability and practicality of Dynamic Range of a DSLR (D700 for example), and some tips on shooting for HDR.

The DXO Mark list D700’s (or D3) reference dynamic range of 12 stops (1 stop = 1 Ev) at ISO 200. So, what does this number mean? This is the maximum dynamic range that the sensor is capable of capturing in a 14 bit RAW image. For all practical purposes, lets assume that really good RAW processing wizards are able to extract about 10 stops of dynamic range from these RAW images. Default settings of most RAW converters (ACR and NX) can get you about 7-8 stops dynamic range in an image.

Technically speaking – the sensor captures plenty DR. More than what typical output devices (printer, monitor etc) can address at this time. In fact, if someone is really very good at RAW processing, one only need to capture an additional +/- 6EV ( D700 already automatically allows one to capture +/- 4EV when using Auto Bracketing with 9 frames) to give a combined 24 EV range. This amount of dynamic range is very close to human eye’s dynamic range!

But then again, there are several limitations on the practicality of this much DR. There are no means to either display or print such high dynamic range images at the moment. So, one pretty much resorts to compressing the tonal range to fit into the medium currently supported. In the process of doing so, we end up exagerrating the details in the shadow and highlights more than what the human eyes can actually see (Thus the “ugly” grungy looking HDRs prevalent on the Flickr! and other popular websites).

Hold on! Lets step back a minute – So, the camera already has the ability to provide 8-10 stops of DR! AND Most  post-processed (a.k.a Tone mapped HDR) images also has the same amount of DR!   So, whats the difference? Why do we need to get into the circus of shooting 3+ images and end up with all that mess with alignment, tone-mapping etc. and still end up with the same amount of dynamic range?

Answer: Whats different between the two is the tonal characteristics of the image.  The ability to play with and change the default curves (ACR, camera curves, EXPEED, BIONZ, DIGIC etc)  is what gives us these variations in tonality (and dynamic range). It is for this reason that shooting RAW and processing the image yourself gives one the edge (defining “the style”).

Fortunately, this above result does not really require one to shoot multiple exposures and merging them using Photomatix. There are several other ways to get to the same results. People have demonstrated this with use of creative use of Photoshop (for example: Andrej Dragan), use of 3rd party filters/add-ons (such as Topaz or Lucis) and several other alternative HDR processing/tone mapping software available (some  free software as well). In any case, extracting the full capabilities of the RAW image can also be accomplished with the use of a very good RAW processor as well. I had a post earlier comparing some good RAW processors here.

In any case,  here are some hows and whys on shooting HDRs without getting very technical (See references for tech speak):

  • A sensor is normally configured by the factory to work in the visible spectrum range only.
  • The RAW file format just captures the data as the sensor sees with very minimal processing.
  • The dynamic range of a typical scene is about 25-30 Ev.
  • The sensor’s dynamic range limit is about 12 Ev.
  • The exposure values determine which part of the spectrum is captured (and how the scene is compressed).

So, here is how you can capture a 24 Ev dynamic range scene:

  • Configure your camera to shoot the highest quality RAW (14 bit NEF in D700).
  • Use Manual exposure; Mount camera on tripod. Measure scene exposure using spot/average, and set exposure in camera.
  • Set White balance and any other parameters you wish (optional, not necessary).
  • Expose one image at average exposure, shoot 2 other images 8 Ev apart. So, that means – if your average scene exposure at ISO 200, f8 was 1/500th sec, the two other exposures would be 1/80 and 1/3200 seconds. (based on the chart here). Note that the -8EV would get very tricky (needing long exposures and associated issues to go with this)  if your average exposure is low to start with (happens if the light conditions of your scene is low). In this case – you will have to bump up the ISO to higher values so that your average exposure is usable and in the 1/250-1/500 range.
  • Once the images are shot and ready to be processed – Import them with your favorite RAW converter – and zero in all the settings for sharpness, linear gamma curve (if you don’t know what this is – just make sure your tone curve is a straight line going across diagonally at 45 degree angle). In Lightroom, you can create a preset with these options and apply them upon import or in batch.
  • Save the individually exposed images as TIFF images (The file actually contains RAW data as well – but is vendor independent). The only reason one would do this is to bypass the generic RAW conversion process available within the HDR tool (such as Photomatix).
  • Import the images in a HDR tool (such as Photomatix or Enfuse) – and have fun creating the HDR image.

I am pretty sure there are advantages to shooting more than 3 shots to cover the same dynamic range of 24 Ev. The benefits/downsides of doing that as compared to capturing just 3 images are the following:

  • More images to merge results in less noise in the final image.
  • More images to merge results in smoother transitions between colors & luminosity in the final image.
  • More images to merge will have more alignment issues.
  • More images to merge requires a higher quality computer to process images (and more time to generate HDR).

I have personally not seen a significant visible difference b/w merging 3 images and merging 9 images to capture the same dynamic range.

Update: Sep 13, 2009:

It appears that there is a bit (or plenty) of critisim on this post on some forums. Like i said before at the beginning of this post, This post is not a tutorial. Creating a tutorial does not require one to know the whys and hows of HDR, and the algorithms involved. I urge people to study these instead of coming up with an yet another “Here is my HDR tutorial”. The post is intended to stimulate thinking and foster some insight and generate discussion into whys and hows of the HDR.

What most people see in Flickr! is not really the true definition and intent of HDR. They are generally looking at ugly, tone-mapped versions of grungy looking grossly distorted, misrepresented unrealistic images. However, I understand and accept that there are several people out there that do this in the interest of Art or Fad. Nothing wrong with it. Just that, this is neither creative nor original these days. It was “original” and creative about 2 years ago when it all started.

Anyways – For those people who consider themselves “experts” on HDR (after reading several books on the topic, or after having done several gallery shows or having run several thousand images through photomatix),  here are a few things for you:

  • I agree (and use it myself) that Photomatix is a good tool to generate HDRs. However, it is its user that makes or breaks an image. The controls within Photomatix is flexible and supports a wide degree of applications – from Forensic to Art to everything in-between.  Photomatix is very unique in its capabilities and generally works better the more pictures you throw at it. In that sense, a 9 frame, 1EV exposure apart stack of images will yield a better results when (only when) using Photomatix. There are other tools out there that don’t work like this, and don’t need so many source images to work with. The reference below provides a link to the list of several HDR programs out there.
  • I urge people to review the links in the references for Zero Noise Photography. This guy (Guillermo Luijk: link to his site is in the references below) is doing some radical development in not only RAW conversion, but also changing the way HDR images have been thought thus far.
  • The best tool out there to produce “real world HDR” is based on an alternative algorithm called “exposure fusion” (see white paper below on this) – and is implemented in Enfuse/Tufuse applications. Generally speaking, these have been providing the “best” realistic looking images. This is very well suited for interior and architectural photography.
  • One of the only books on the topic i recommend is the”HDRI Handbook” by Christian Bloch. This is a very technical book, and is the reference standard. All others are just “tutorials” in my world.  No better than this.

References:

  1. http://www.cis.rit.edu/people/faculty/ferwerda/publications/EGSR05-hdr.pdf (Tone mapping white paper)
  2. http://research.edm.uhasselt.be/~tmertens/papers/exposure_fusion_reduced.pdf (Exposure blending -Enfuse algorithm).
  3. http://www.hdrsoft.com/resources/dri.html (HDR FAQ from Photomatix guys).
  4. http://www.maxmax.com/hot_rod_visible.htm
  5. http://en.wikipedia.org/wiki/Anti-aliasing_filter
  6. http://www.maxmax.com/nikon_d700_study.htm
  7. http://www.naturfotograf.com/D3/D3_rev05IR.html
  8. http://static.photo.net/attachments/bboard/009/009XXU-19705684.pdf
  9. http://luminous-landscape.com/forum/index.php?showtopic=17775
  10. http://www.guillermoluijk.com/article/nonoise/index_en.htm
  11. http://luminous-landscape.com/forum/index.php?showtopic=37060&hl=photomatix
  12. http://www.hdrlabs.com/tools/links.html

197438 pages viewed, 514 today
32129 visits, 46 today
FireStats icon Powered by FireStats

8th Cross is Digg proof thanks to caching by WP Super Cache