Pages: [1]   Go Down

Author Topic: Exposure Fusion  (Read 2144 times)

rgs

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 603
    • Richard Smith Photography
Exposure Fusion
« on: May 24, 2014, 01:32:14 pm »

I have seldom liked the look of HDR - especially those that are very flat with shadows that are much too light and those that look more like an illustration than a photograph. But I have been working a bit with Exposure Fusion, primarily with the Enfuse LR plugin. Doing some real estate photography, I have had some very good results with interiors and EF.

Today one (an exterior shot with some deep shadows and a bright sky) turned out very flat. I was able to get a much better result with a single exposure and some work in LR. So I am thinking of why the EF result was not successful and I want to run this past all you experts to see if I understand it. Please feel free to comment (or correct).

In general I'm thinking that HDR extrapolates results (isn't that kind of what tone mapping does) while EF merely automates (in a very sophisticated way) the kind of PS blending techniques we've all done by hand. Is that a good simplified view of the two? If I'm right, EF can not produce anything that is not in one of the original exposures. So, with EF, it is good to have 5 (or more) exposures differing by about 2 stops each. That would give the software plenty of data to work with. Today's failure probably just didn't have enough dynamic range for EF to be effective.

Am I on the right track? 
Logged

mdijb

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 666
    • mdiimaging.com
Re: Exposure Fusion
« Reply #1 on: May 24, 2014, 02:49:13 pm »

I also do not like the Grunge look that  tonemapping often produces.

I have had much better results using PHotomatix--use the fusion option, not the tone mapping option.  Select the "Natural" preset.  You may make some tweaks here, but i do not.  Then send it back to LR or Photoshop and process from there.  You will get very good images that look natural.

You also may try using Photshop HDR module.  this creates a 32 bit result.  Send this image back to Photoshop or LR with no adjustments--then process from here.  Again you will get natural looking results

MDIJB
Logged
mdiimaging.com

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Exposure Fusion
« Reply #2 on: May 25, 2014, 08:43:42 am »

Am I on the right track?  

Hi Richard,

It's a bit different. An HDR file has a huge dynamic range, which will render as having very low contrast when you fit all that range into a small range of, say, 0...255. To get a more lifelike rendering, the tonal range of the HDR file will need to be tonemapped, i.e. apply locally adjusted contrast boosts and reductions. HDR files typically have linear gamma encoded data, but can also use floating-point numbers to encode the huge range. The tonemapping requires a lot of effort to keep looking natural.

Enfuse (exposure fusion) does something else, it basically uses selective (based on luminance) parts of already gamma adjusted images and adjusts the brightness and then blends the images together. Blending already gamma adjusted natural looking image fragments, tends to retain a natural look much easier.

Exposure Fusion of images that are exposed more than some 1.33 stops apart, may cause issues in noise differences between different parts of smooth gradients.

Cheers,
Bart
« Last Edit: May 25, 2014, 08:46:33 am by BartvanderWolf »
Logged
== If you do what you did, you'll get what you got. ==

rgs

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 603
    • Richard Smith Photography
Re: Exposure Fusion
« Reply #3 on: May 25, 2014, 07:44:04 pm »

I think that's similar to what I was imagining in my very untechnical way. I appreciate the head up on the 1.3 stop difference. I have played with 1 and 2 stop brackets and generally like 2 stop better. Most of the resources I have found recommend 2 stop brackets. I'll give 1.3 a try. Maybe it's the sweet spot. I think one of the keys is to get a wide enough range between the lightest and the darkest bracket so the software has good data at the extremes to work with.
Logged

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Exposure Fusion
« Reply #4 on: May 26, 2014, 05:44:23 am »

I think one of the keys is to get a wide enough range between the lightest and the darkest bracket so the software has good data at the extremes to work with.

Hi Richard,

Correct. Using 1.3 stops as a maximum interval obviously increases the number of bracketed exposures to cover a given range compared to a 2 stop interval. One can of course also mix this with noise reduction software, which would allow to use larger increments, but noise reduction also risks losing a bit of resolution. Cameras with very low noise may be slightly more tolerant.

I arrived at the maximum interval of 1.3 stops after comparing a lot of image fusion results. I made a huge series of 1/3rd stop intervals (at ISO 100) of a high dynamic range scene (partly in the sun, partly in the shadow), and then compared all combinations. First fusing all 0.3 stops brackets, then all 0.6 stop brackets by leaving out every other image, then all 1.0 stop intervals by skipping 2 brackets, and so on. I then noticed that at certain smooth gradients it became obvious that the smoothness of the gradient was negatively impacted by a too large (>1.3 stops) interval.

When the steps get too large, it may also become harder to align the images before fusion. Alignment is necessary for handheld bracketed series, but may also be needed for longer focal length images. A single pixel offset of one or more of the images, can influence the resulting fusion.

I used to use Tufuse Pro (similar to Enfuse), but prefer SNS-HDR (also uses exposure fusion, not HDR merge) because of it's virtually halo free tone-mapping quality.  

Cheers,
Bart
Logged
== If you do what you did, you'll get what you got. ==
Pages: [1]   Go Up