|
Convenient web based access to our favorite web design Usenet groups
|
 |
This is Interesting: Free Magazines for Graphics designers and webmasters
| Author |
| Thread |
 |
|
|
|
|
|
|
|
|
|
|
 09-15-04 - 05:14 PM 
|
In article <41478064$0$20254$cc9e4d1f@news-text.dial.pipex.com>,
Bobb_Dogg@hotmail.com says...
>
>When I scan in with my Nikon IV I get a smooth histogram.
>
>If I adjust Levels etc, "teeth" appear in the histogram with gaps and =
>spikes. Does this matter for print quality using an Epson 890? I've =
>read somehwere that spikes like this mean posteization. Is that right?
>
>More importnat, is there any way I can flatten these spikes out after =
>i've adjusted Levels and curves, so I get a better image but no =
>posterization?
Why? The histogram gaps are normally not an issue, normally not visually
perceivable in the image itself. We can only see this effect when viewing
the histogram. These gaps and spikes are caused by adjusting (shifting)
the tone values in 8 bit data with the Level, Curve, or Color Balance
tools. When you shift the tone values, say those tones at value 200, up to
now be value 202 (just to have a number), in some cases you might leave a
gap at 200, only meaning there are no data values remaining with value 200
(no other values to be shifted to become 200). If there is a spike at 202,
it just means the count combined with other tones already there, already
with value 202. The histogram simply shows the count of pixels for each
possible tone value. The only meaning of the gap is there are no pixels
with tone value 200, and the only meaning of the spike is that perhaps
there are double the number of pixels with tone value 202. Generally,
thee main idea is that for full image contrast, we adjust so that the
histogram data extends full range, from 0 to 255. There is no law that
says the real scene image data must have values at every possible data
value, and there is no law that says we must have an equal tonal
distributions (the proverbial image of black cat in coal mine, or polar
bear in snow storm, are certainly big exceptions to everything). The
histogram pixel counts simply depend on the image content, and images
differ from each other. More importantly, the human eye can barely
distinguish values 200 from 202, so this gap/spike is certainly no big
deal. You wanted to change the tone values, remember?
If this is perceived as a problem, the solution is that we can make such
tonal changes (at least the drastic changes) in 16 bit data, so that the
gaps/spikes are unseen. This means to output 16 bit data from scanner,
change the tones in 16 bit data in Photoshop, then convert to 8 bit mode
afterwards. Make no mistake, we still get the same gaps/spikes in 16 bit
data, but we just cannot see those gaps/spikes when the 16 bit data is
displayed in a 8 bit histogram display, or after the data is
regrouped as 8 bit mode. Many of us do it this way, and for that reason,
but there is no actual difference in results when viewing the image.
Otherwise, if ignored, in very worst case, the nit-pickers remind us that
the gaps and spikes might possibly theoretically combine in an extreme way
so that the most extreme cases of wide gaps might possibly might become
visible in the image, seen as banding (posterization) in very wide smooth
gradients (sky, walls, etc). But as a general practical case, it simply is
not an issue. We look at the image, not the histogram.
--
Wayne
http://www.scantips.com "A few scanning tips"
|
|
|
| [
Post Follow-Up to this message ]
|
|
|
|
|
|
|
|
|
|
09-16-04 - 12:14 AM
|
Wayne
Thanks for taking the trouble to provde such a long and lucid explanantion
I'll not lose too much sleep and go on with the correction methods i use.
Bobby D
"Wayne Fulton" <nospam@invalid.com> wrote in message
news:TN6dnZVXRfgjyNXcRVn-tg@august.net...
> In article <41478064$0$20254$cc9e4d1f@news-text.dial.pipex.com>,
> Bobb_Dogg@hotmail.com says...
>
>
> Why? The histogram gaps are normally not an issue, normally not visually
> perceivable in the image itself. We can only see this effect when viewing
> the histogram. These gaps and spikes are caused by adjusting (shifting)
> the tone values in 8 bit data with the Level, Curve, or Color Balance
> tools. When you shift the tone values, say those tones at value 200, up to
> now be value 202 (just to have a number), in some cases you might leave a
> gap at 200, only meaning there are no data values remaining with value 200
> (no other values to be shifted to become 200). If there is a spike at
> 202,
> it just means the count combined with other tones already there, already
> with value 202. The histogram simply shows the count of pixels for each
> possible tone value. The only meaning of the gap is there are no pixels
> with tone value 200, and the only meaning of the spike is that perhaps
> there are double the number of pixels with tone value 202. Generally,
> thee main idea is that for full image contrast, we adjust so that the
> histogram data extends full range, from 0 to 255. There is no law that
> says the real scene image data must have values at every possible data
> value, and there is no law that says we must have an equal tonal
> distributions (the proverbial image of black cat in coal mine, or polar
> bear in snow storm, are certainly big exceptions to everything). The
> histogram pixel counts simply depend on the image content, and images
> differ from each other. More importantly, the human eye can barely
> distinguish values 200 from 202, so this gap/spike is certainly no big
> deal. You wanted to change the tone values, remember?
>
> If this is perceived as a problem, the solution is that we can make such
> tonal changes (at least the drastic changes) in 16 bit data, so that the
> gaps/spikes are unseen. This means to output 16 bit data from scanner,
> change the tones in 16 bit data in Photoshop, then convert to 8 bit mode
> afterwards. Make no mistake, we still get the same gaps/spikes in 16 bit
> data, but we just cannot see those gaps/spikes when the 16 bit data is
> displayed in a 8 bit histogram display, or after the data is
> regrouped as 8 bit mode. Many of us do it this way, and for that reason,
> but there is no actual difference in results when viewing the image.
>
> Otherwise, if ignored, in very worst case, the nit-pickers remind us that
> the gaps and spikes might possibly theoretically combine in an extreme way
> so that the most extreme cases of wide gaps might possibly might become
> visible in the image, seen as banding (posterization) in very wide smooth
> gradients (sky, walls, etc). But as a general practical case, it simply is
> not an issue. We look at the image, not the histogram.
>
> --
> Wayne
> http://www.scantips.com "A few scanning tips"
>
|
|
|
| [
Post Follow-Up to this message ]
|
|
|
|
|
|
| All times are GMT. The time now is 05:26 PM. |
|
|
|
|
|
| |
|
