Olympus-OM
[Top] [All Lists]

[OM] Re: Ingrid Bergman - two new images

Subject: [OM] Re: Ingrid Bergman - two new images
From: Moose <olymoose@xxxxxxxxx>
Date: Sun, 19 Nov 2006 01:56:27 -0800
Winsor Crosby wrote:
> I was just thinking about the overexposure bit. Maybe there is more  
> to exposure than I thought. While it is true that it is not truly  
> overexposed to the point of the red channel being slammed up against  
> the right boundary of the histogram it is the only channel with any  
> information to the right of the center line. Both the blue and the  
> green channel stop near the center line. 
Think of a pic of a card all one color red. The histogram for the other 
two channels will show little or nothing, depending on how close the red 
is to the red of the sensor design and how good the filters on the 
pixels are. The red histogram should theoretically show a single 
vertical bar. Again there will be some imperfections resulting in 
something like a tight bell curve.

That may look totally wrong, if you are used to thinking of histograms 
that are nice bell curves and don't look at the separate channels very 
often. However, it is a pretty accurate representation of the subject 
and will print a red card much like the original subject.

With this subject, I would expect a histogram more or less like what it 
has. Almost all of the higher brightness tones are variations of 
something close to pure red, so most of the histogram above the center 
line SHOULD be red, otherwise, the rose would be a different color.
> So the rose is just shades  
> of red - little blue or green except for the deepest shadows. If you  
> do anything to increase the red value it lightens the darker red that  
> is the detail but is still to the right of the midline and it goes  
> all flat. It has the effect of overexposure. If you look at the  
> overall histogram of the corrected image I sent Brian and move the  
> right hand arrow to the left you can watch the detail disappear in  
> the rose.
>   
Moving the right hand arrow is only for bringing the highest values in 
the histogram up to the end if they aren't already there (or for 
intentionally blowing out highlight detail for artistic reasons.) If you 
want to raise the overall brightness of the rose using the Levels conrol 
without losing highlight detail, move the center slider of the red 
channel to the left. (You REALLY want to be working in 16 bit here.) 
That stretches out the tonal values to the left of the existing center 
point and compresses those to the right. It moves all the red tones up 
except at the ends.

The problem with this approach is that it changes the color of the 
leaves and shadows. If you really want to do it, and not throw off the 
color, you need to select the blossom and only work on that.

I don't see why this would be necessary. Curves can easily bring up the 
lighter tones, most of which are red anyway, and lighten the bloom. 
Messing with the color channels in Levels and Curves WILL change the 
color balance.

The histogram of the original shot looks pretty decent to me, but the 
image look pretty flat and soft. LCE takes care of some of the flat look 
and some of the unsharp look. Curves takes care of the rest of the dull 
look, brightening up the center of interest. Sharpening takes care of 
the rest of the softness. With this image, it's first  necessary to use 
Contrast and/or Brightness, depending on whether you care about the 
shadows, to bring the top of the histogram to the left to leave room for 
the effect of LCE. Otherwise, the highlights will blow out and change color.

I think the version I posted before nicely brings up the bloom without 
messing with color balance. If the way it also brings up the largely OOF 
background doesn't work, you really need to get it into a separate layer 
to adjust it, otherwise color balance changes in all the green areas.

I've added another alternate to my version with lowered background 
brightness and some adjustment to the foreground leaves 
<http://www.moosemystic.net/Gallery/Others/BSwale/IBergman.htm>.

Moose

==============================================
List usage info:     http://www.zuikoholic.com
List nannies:        olympusadmin@xxxxxxxxxx
==============================================

<Prev in Thread] Current Thread [Next in Thread>
Sponsored by Tako
Impressum | Datenschutz