DoctorChas,

To attempt to explain this in a slightly different way I will use an analogy that works for two similar imaging devices, video, and digital single frame cameras.

If you were to go outside on a bright sunny day and view a scene with normal Human eyes you would notice a very nice photographic opportunity. If you were to measure the light throughout this scene you would have a very wide range of light measurements.

If you now take your camera and use its exposure meter to measure the same scene you may well record a very wide exposure range as the camera passes over the bright sun lit areas, and the shadowed areas. Let us propose your camera can capture an exposure range of +1 - to -1 Ev from the average of the scene at 0Ev.

Now let us shift this measurement over to video. +1 Ev = .7 volts above 0v. (the average), and -1Ev = 0v. (Total black shadows). When the digital sensor reaches .7 volts the captured image blows out to pure white, and when the digital sensor falls to 0v. the captured image descends to pure black. (This holds true for Color and B&W, this is the Luminance Channel.)

When you set your camera up for bracketed captures, and each exposure is set for +1.75 Ev, 0 Ev, and -1.75 Ev, you have extended the exposure range of the sensor to well beyond .7v. to below 0v (-.3v). With the additional two exposures.

What you will accomplish is taking that part of the image, under normal exposure values, that would be clipped at .7volts and lowering the exposure value so that all the clipped areas would fall below .7volts. You would, also, take those parts of the scene that fell below 0v and raise the exposure values above 0v.

When you transfer your exposures into your computer and apply HDR Processing, you merge all three exposures into one well balanced final image. (The merging is usually performed by software using Masks and Layers to blend all three image exposures.) HDR captures is usually best made by setting the camera to Aperture Priority, this causes the Shutter Duration to make the exposure Value changes.

Michael G

Sorry Doctor, you can rattle off all the numbers you like but even if you have everyone else in awe of your apparent knowledge, you are wrong. I will say it again....bit depth has nothing to do with the actual dynamic range to which any given sensor can usefully respond. Bit depth refers to the degree of resolution of the recorded data. It looks like I might have to write another one of my (annoying to some) threads to properly explain this concept to people who have it hopelessly screwed up. Cheers, Rob.

Just for kicks, what happens if you take that HDR image that is created and don't tone map it at all. Does the program allow you to save it as a 16 bit file? or can you save it as 32bit and then use PS to save it as 16 bit?

Does that 16 bit image then look like the raw file you started out with? I tried something similar a while back to what you are doing with PS and I couldn't make the tonemapped image significantly different from the initial raw. I guess I never liked the PS tonemapper and ended up with Photomatix.

The images are not functionally equivalent. For the two you posted, if you look at the shadow detail under the pallet the ropes sit on you can see that the fake image is not as clean as the 3 shot real image. Not a lot of difference but it does exist. Now what you are trying to do works if you take the fake image and process it to where you like it and then take the real image (3 shot data) and try to mimic the fake image. You'll be able to do that because the 3 shot data incorporates the dynamic range of the single shot fake image.

But if you tonemap the 3 shot data real image to the point where you've brought out the shadow and highlight detail to their maximums before creating issues such as noise, there is no way that you can duplicate that with a fake one image shot.

If you try another image as a test that has a greater dynamic range, you can prove this to yourself. The one you use as an example just doesn't have enough range to prove the point. Again, I understand the point and desirablity of this method in certain circumstances, but they are not functionally equilvalent.

First HDR is High Dynamic Range.
RAW creates images of up to 6 step in the human vision that is effectively limited to about 24 for the best and 14 or even lower for many.
JPG captures only 2 steps.
When shooting in RAW you have a potential of 6 steps (but usually reduced to 4). You can create what is called a 'pseudo HDR image' by playing with levels and curves with a single RAW file.
A JPG will never be able to that and requires at least 3 images if not more.
The key here is to understand that to get a 'true' HDR, one has to be aware that you are not dealing with color or color shade but luminosity. Color depth is not an issue here, not yet anyway.
When shooting for HDR using multiple images one has to be aware of the true format and camera capabilities.
When processing these images one MUST set the image bit depth to 16. Most software defaults to 8. This is true, especially for JPG that is limited to 8 by default.
Why then? Simply because when you manipulate the image and blending them you take advantage of the software superiority. When you save the image back to JPG it is reduced to 8 bit but with NEW data*...
I am sure I am not explaining this well enough, if someone can do it better, please do so.

* This data is created by the higher bit depth and, in this case, the blending. It somehow corrects some of the limitations of the JPG bit depth format. This is true for ALL JPG manipulations by the way.

Not to steal the thunder, but I believe Photomatix has an option to load a raw file and then tone map it. You can also batch process single files. I don't know if the hdr program you use has that capability.

Tonemapping a single image (or brackets taken thereof) and changing the image in LR or Aperture is essentially the same since you are just working with a raw file that has the ability to move 2 stops under or over the taken exposure. You can go further but then you have noise, etc. The difference between doing it in camera is that if you take a stop and a half over and under that really gives you an additional three and a half over and under to work with during the tonemapping process.

Pulling brackets from a single image doesn't get you more range it just allows the tone mapping algorithims to be applied to the raw file - that's not available in LR for instance when you are just executing image adjustment - you won't get that hdr look. But I get the argument about moving subjects where the ghost removal becomes problematic and this is a way around it.

For the example you used, it works but if you had an image with a large dynamic range, it probably wouldn't come out as well as if you did true brackets. You could probably move the histogram for the 3 brackets in camera around much further and still have quality that you won't get by working with a single file?