vuescan and clipping whites - any help graciously accepted

[Figo]

Hello group

Scanning noobie here looking for some help. I have waded through two
books but am still unclear on a few matters. Perhaps someone who is
familiar with vuescan could answer this best.

All of vuescan's colour defaults (manual, neutral, white point, etc)
come with a suggested white clipping that is higher than 0%. Black point
percentage is 0%, but white point percentage is 1%. By my understanding,
that necessarily means that that you lose some hightlight detail.
Everything in the last percentile of the highlights will be blown out.

My questions (misunderstandings?) are two fold.

First of all, this default of 1% looks a hell of lot more than 1% on
their histogram graph. The little triangle beneath the histogram looks
like it's moved in to about 10% rather than 1%. Why is this so?

Secondly, I do not understand why you'd want to clip any white at all?
Surely by blowing out the highlights you are losing detail in the
process, which is critical if I'm scanning for printed output?

I am a bit lost in all of this - any help will be graciously accepted.

 

[Jeff Randall]

First of all, this default of 1% looks a hell of lot more than 1% on
their histogram graph. The little triangle beneath the histogram looks
like it's moved in to about 10% rather than 1%. Why is this so?

1% represents the percentage of total image pixels clipped not
distance on the X-axis grayscale.

Secondly, I do not understand why you'd want to clip any white at all?

Experiment. However, many images will look flat (ie., not have
sufficient contrast) without some white point clipping. Typical
consumer photo lab snapshots often clip as much as 5%. You can of
course scan with the black and white points set to 0% (as many folks
do) and then tweak the image with curves in a photo editor.

Jeff Randall

 

[Kennedy McEwen]

Hello group

Scanning noobie here looking for some help. I have waded through two
books but am still unclear on a few matters. Perhaps someone who is
familiar with vuescan could answer this best.

All of vuescan's colour defaults (manual, neutral, white point, etc)
come with a suggested white clipping that is higher than 0%. Black point
percentage is 0%, but white point percentage is 1%. By my understanding,
that necessarily means that that you lose some hightlight detail.
Everything in the last percentile of the highlights will be blown out.

My questions (misunderstandings?) are two fold.

First of all, this default of 1% looks a hell of lot more than 1% on
their histogram graph. The little triangle beneath the histogram looks
like it's moved in to about 10% rather than 1%. Why is this so?

1% of the total number of pixels, not 1% of the peak white level.
Depending on the histogram shape, 1% of pixels may well be 10% or more
lower in level than the very brightest pixel in the image.

Secondly, I do not understand why you'd want to clip any white at all?
Surely by blowing out the highlights you are losing detail in the
process, which is critical if I'm scanning for printed output?

There are lots of reasons for clipping both white and black. Probably
the most obvious is that the peak may not even be a real part of the
image, but could be dirt, dust or even damage to the emulsion, and you
don't want that determining the contrast of your image. Another reason
for clipping white, but not black, comes down to noise - there is just
more noise in the highlights than there is in the shadows (for a slide
anyway - see the thread on multisampling for an explanation of this)
and, again, you don't want noise to drive the contrast setting for the
image.

 

[JB]

Thank you for answering.

1% represents the percentage of total image pixels clipped not
distance on the X-axis grayscale.

Okay, starting to vaguely understand now.

So this setting means that all pixels in the brightest 1% are clipped,
regardless of their values, rather than the pixels that fall into the last
percentile of brightness? (which would be about 253-256 on an image with a
nice bell shaped histogram that kisses both book ends)

Correct, or am I heading off in the wrong direction here?

 

[Mendel Leisk]

Hello group

Scanning noobie here looking for some help. I have waded through two
books but am still unclear on a few matters. Perhaps someone who is
familiar with vuescan could answer this best.

All of vuescan's colour defaults (manual, neutral, white point, etc)
come with a suggested white clipping that is higher than 0%. Black point
percentage is 0%, but white point percentage is 1%. By my understanding,
that necessarily means that that you lose some hightlight detail.
Everything in the last percentile of the highlights will be blown out.

My questions (misunderstandings?) are two fold.

First of all, this default of 1% looks a hell of lot more than 1% on
their histogram graph. The little triangle beneath the histogram looks
like it's moved in to about 10% rather than 1%. Why is this so?

Secondly, I do not understand why you'd want to clip any white at all?
Surely by blowing out the highlights you are losing detail in the
process, which is critical if I'm scanning for printed output?

I am a bit lost in all of this - any help will be graciously accepted.

I also find this 1% very high for "default". For tri-x scans, I prefer
..02% off both ends, which I find just cleans up the stragglers at
either end of the histogram. With the few slide scans I've done, 0%
both ends seems fine.