Jeff Coatney
Well-known member
I have a question about HDRx™.
Let's say you take an HDRx™ shot where the under is perfectly exposing the darker parts of the image but not clipping the highlights, just overexposing them; and then the over is exposing the highlights perfectly but not clipping the shadows, just underexposing them. In this situation HDRx™ makes perfect sense to me as the blend should create an image where everything is properly exposed.
Now let's say you take an HDRx™ shot where the shadows and highlights are further apart. The under is perfectly exposing the shadows but the highlights are significantly clipped; the over is perfectly exposing the highlights but the shadows are significantly clipped. I figure this is pretty typical. So what I don't understand is how this would have the same effect as in the previous example… because now the clipped shadows, whether they were barely out of the exposure range or very far out of the exposure range, are blending in as if they were all the same color (black); and the clipped highlights, whether they were barely out of the exposure range or very far out of the exposure range, are blending in as if they were all the same color (white). So now any part of the blended image that was blended with either clipped shadows or clipped highlights is still getting "half clipped" if that makes any sense.
For instance, let's say a wall is filling the frame, and the wall is being lit in such a way that one side is in shadow and the other side is brightly lit, and it gradually changes from one to the other so that the middle of the wall is 50% as dark as the shadow side and 50% as bright as the well-lit side (so the wall is a linear gradient). Using HDRx™, wouldn't the resulting blend have a consistent, linear ramp up from the dark to light in the middle of the wall where neither the under or the over was clipping, but then once you get to far enough in either direction that either shadows or highlights are clipping, the change in apparent exposure would no longer account for the clipped values?
If that doesn't make any sense, think about it like this:
100 = white
000 = black
1XX = clipped highlights
(assume the pattern remains the same over this imaginary gradient)
OVER|UNDER: BLEND
020|069: 44
026|075: 50
032|081: 56
038|087: 62
044|093: 68
050|099: 74
056|1XX: 77.5 ("should be" 80)
062|1XX: 80.5 ("should be" 86)
068|1XX: 83.5 ("should be" 92)
Hopefully that makes sense. As you can see (?), once you get to the part of the gradient where there is clipping in either the under or over image, the blend becomes increasingly inaccurate.
So my question is: does HDRx™ take note of what is being clipped (whether it be in the under or the over) and compensate in the blend by replacing clipped values with "super white" or "super black" values which are calculated using the exposure metadata and values from the other image (under or over) in which that part of the image was not clipped?
I'm not trying to figure out how you've done what you've done from a technical standpoint (I am interested, but I know you have to keep it close to your chest), but rather, I'm trying to determine the limitations of HDRx™. I'm assuming you do somehow compensate for this issue, but if not… couldn't you? Using a system like the one I described in the previous paragraph?
David, I think the point is to expose as accurately as possible and then let HDRx work its magic. In the example you present, if I'm reading it correctly, an error in either direction becomes uncorrectable. I would assume this to be the case, given the whole point of HDRx is highlight protection. The example of the wall you give should have an acceptable exposure range if the entire shot is correctly exposed-- it being exposed in the context of the overall scene. But can the result be skewed? Perhaps on purpose? I'd say yes. And that just adds to the creative options. The compensation you seek is in your own creative choices, IMO.