Re: Ready to graduate to DSLR

Chris Brown wrote:
> Bryan Olson wrote:
>
[Chris Brown wrote:]
>>>It's there in black-and-white - your claim is specifically that "digital
>>>images are far more flexible". Since digital images are new, it clearly
>>>follows that you believe this greater flexibility, apparently inherent to
>>>digital images, is also new.
>>
>>Why not interpret it to mean what it actually says. The flexibility
>>means this particular rule -- to expose at a level for final viewing,
>>no longer holds. If you had a different process so the rule didn't
>>apply anyway, great.
>
> But you specifically said it was a flexibility of *digital* images. Is it,
> or isn't it?

Yes, of course, just as I wrote. I've lost track of what you are arguing.

>>>So if you don't think this "exposing to the right" is, after all, a new
>>>technique enabled by the glorious histogram, then did you actually have a
>>>point at all?
>>
>>Histogram display is a winner.
>
> How so? I think histograms are great, but not because they enable
> highlight-oriented exposing, or even make it easier, because they don't.

Not sure, but I think I see the problem. The point I was making about
decoupling exposure at data-capture from exposure in the final image, and
the point of the expose-to-the-right-technique, is not that one should
never let a highlight blow out.

In most cases, one wants the saturation point of the sensor to coincide
with a low-point, a valley, a basin, of the image's histogram. Digital
sensors (perhaps with rare exceptions) do not degrade gracefully at the
high end. When transitions that should look smooth go from showing detail
to blowing out, the change is often noticeable and, when it is noticeable,
is almost always undesirable.

So when, as usual, the saturation point falls in histogram valley, what
is the effect of allowing more light -- moving the histogram to the
right as far as is possible -- while keeping the saturation point in the
trough? The positive effect is a better signal-to-noise ratio. Obviously
there may be other ancillary effects from the change that admitted more
light, be it shutter speed, aperture, or additional lighting, but those
are not the issue here. There is no general negative effect of
significance.

"Expose to the right" is not exactly the same thing as "highlight-oriented
exposing". Expose-to-the-right is not really about moving histograms to the
left, though there are cases where the histogram can alert one to problems
that should be solved by admitting less light. It is about capturing the
most accurate and precise data possible, under the assumption that as long
as the data is good, subsequent rendering can scale values as needed for a
visually appealing result.


--
--Bryan
.

Relevant Pages

  • Re: Fast search of images
    ... > To this we have the costant that is 256 if we use the histogram + the cost ... >>> I'm tring to write a code to search for similar images in a database. ... >>> checking in a binary tree to performe the search in O. ... You can always store histograms with various numbers of bins. ...
    (sci.image.processing)
  • Re: Histogram for color images
    ... title('Original Color Image', 'Fontsize', fontSize); ... I have run the demo and later i want to run it whith other images. ... I have seen a paper entitled " Key frame extraction from MPEG video stream" and they mentioned the the difference color histogram between two frame color images as follows: ...
    (comp.soft-sys.matlab)
  • Re: Striking tonal range in B&W images - How?
    ... bathtub shaped histogram where they are very black and very white ... My images are very gray with a fairly flat and constant histgram. ... My images end up looking for flat and gray similar to AP51F2 and ARF350 ... JPEG and high noise ...
    (rec.photo.digital)
  • Re: Fast search of images
    ... To this we have the costant that is 256 if we use the histogram + the cost ... >> I'm tring to write a code to search for similar images in a database. ... >> Using the histogram to perform a faster search works but still the ...
    (sci.image.processing)