Pages: [1]   Go Down

Author Topic: trying to understand saturation  (Read 11757 times)

sgwrx

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 310
trying to understand saturation
« on: May 29, 2006, 01:06:53 pm »

i'm trying to understand saturation. i noticed that increasing saturation in the extreme has the effect of lowering the height of a histogram. if i remember correctly, the histogram height was the count of the number of pixels present at each point along the horizontal axis. this seems to indicate that increase saturation actually results in lost data, ie a small area that had 20 different shades of yellow is reduced to 5 shades or 1 shade of solid yellow.

next i whipped up a document and painted several different squares of color (r, y, g, .

the first set of squares were: r - 255,0,0 y - 255,255,0 g - 0,255,0 b 0,0,255. i used the hsl slider to increase saturation and nothing changed in appearance or the rgb values.

the second set of squares were: the same as above but i set the saturation in the hsl color picker tool to 50%. again i increased saturation and nothing happened. i decreased saturation and the rgb values all went down at different rate until they all became 191,191,191.

finally, i played around with some arbitrary versions of red, green, blue and yellow squares. increasing and decreasing saturation.

it seems that for a given color, increasing the saturation will work to increase the rgb values that make up that color until one of the 3 values reaches 0 and no further change will happen. further, lightness plays an important role. if lightness is adjusted up, then saturation increased, all rgb values increase until one of the values reaches 255 and then no further changes happen.

it all seems quite complicated.

what are some general guidelines that can be had from this when trying to balance loss of tonality vs increasing saturation?

ps - any books specfically on color saturation? i have Color Mgmt for Photographers, Mastering Digital Color (and B&W) Photography.

thanks
Logged

dlashier

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 518
    • http://www.lashier.com/
trying to understand saturation
« Reply #1 on: May 29, 2006, 01:15:23 pm »

Saturation has to do with the spread between r, g, and b values. For instance 30,60,90 will be more saturated than 40,60,80 or 50,60,70, and of course, 60,60,60 is totally desaturated. This is why saturated colors are most susceptible to (channel) clipping or falling out of gamut.

- DL
Logged

PeterLange

  • Guest
trying to understand saturation
« Reply #2 on: May 29, 2006, 06:08:22 pm »

Quote
i'm trying to understand saturation....

There are quite different (math) definitions. The saturation slider of Hue/Sat.-tool in Photoshop can operate in at least three different ways:

1.) The Hue&Sat.-saturation slider in RGB mode (Normal blending) combines HSB-saturation with HSB-brightness. In detail, a pure increase of HSB-saturation compares to a ‘spectral purification’ where the less dominant wavelengths (resp. lower RGB values) are further reduced. So it’s not wrong in principle to add some brightness in order to avoid a ‘too dark’ problem.
http://en.wikipedia.org/wiki/HSV_color_space

2.) The Hue&Sat.-saturation slider in RGB mode (Saturation or Color blending) switches to a proprietary version of the HLS color model, wherein Luminosity is computed as a weighted average of green, red and blue. The whole architecture finally accounts for many similarities with Lab; though it’s still a RGB derivative and can not be exactly the same as Lab. However, at moderate settings, the ‘perceived’ Lab-lightness of most colors is nearly maintained, and the HSB hue is always perfectly maintained.
- my favorite -

3.) The Hue&Sat.-saturation slider in Lab mode (Normal blending) joins the Lab / Lch definition of color saturation. Colors are moved along lines of constant a-b-angle, straight away from the inner Lab-lightness axis (for an increase of saturation). Finally, that’s right the same as this ‘new’ technique to steepen the contrast on both channels a&b. Further, without intending to overemphasize this point, changes of Lab-saturation bear some in-built distortions of the perceived hue & perceived saturation as defined e.g. by Munsell colors. These ‘design flaws’ of CIE Lab are described in great detail on Bruce Lindbloom’s website (see his concept for a Uniform Perceptual Lab and further links).

http://www.outbackphoto.com/tforum/viewtop...picID=1657#8085

Further options can be realized via the Channel Mixer, or the assignment of chroma variants of color spaces…

Peter

--
Logged

sgwrx

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 310
trying to understand saturation
« Reply #3 on: May 29, 2006, 08:34:20 pm »

thanks both.  good things for me to look up.

with regard to the channel clipping, i did some playing around earlier and found these things:

as long as one of the colors contains one or two RGB value of 0, then increasing saturation does nothing.

if one or two RGB values is 255, then increasing saturation does nothing.

increasing saturation can lead to individual R G or B clipping or maxing out at 255.  if that happens, then increasing saturation any more, does nothing.

i'm going to play around in lab to see what's up with that. again thanks.
Logged

sgwrx

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 310
trying to understand saturation
« Reply #4 on: May 29, 2006, 08:41:08 pm »

interesting:

i started with a LAB file, did a spectrum gradiant. then converted to RGB color and desaturated with HSL layer. i got bands of gray vs. starting wth an RGB file and spectrum then desaturating with HSL layer.

this is getting interesting.
Logged

PeterLange

  • Guest
trying to understand saturation
« Reply #5 on: May 30, 2006, 05:12:51 pm »

Quote
with regard to the channel clipping, i did some playing around earlier and found these things:

as long as one of the colors contains one or two RGB value of 0, then increasing saturation does nothing.

if one or two RGB values is 255, then increasing saturation does nothing.

increasing saturation can lead to individual R G or B clipping or maxing out at 255.  if that happens, then increasing saturation any more, does nothing.

That’s how it is.

Finally, the Hue/Sat.-tool lets the data bump against the border of your RGB working space which is indicated by R or G or B= 0 or 255.  When this happens with too many pixels, posterization can occur.  Possible solutions:

/a)  To convert to a larger working space like ProPhoto RGB which offers more headroom.  That way, there’s of course a certain risk to overdo it; to create fancy saturated colors (or even imaginary colors) which lead to output-referred problems (monitor / print). So it can be wise to combine this with a Selection e.g. as suggested in point d).

/b)  Saturation Curves, e.g. such as offered by the Curvemeister plug-in.

/c)  To convert to Lab mode and to work with S-shaped curves on both color channels, a & b.  Probably great, but not really my field of competence.

/d)  Saturation Masks.  Just recently, I’ve learned & implemented the following technique:

/>  Create a Selective Color layer. For all R, Y, G, C, B, M color ranges set Black to -100%, respectively. For the White, Gray, Black ranges set Black to +100%, respectively (everything Absolute).
/> Strg + click on the RGB composite channel (to select everything visible)
/> Strg + Shift + I to Invert the selection
/> Delete the Selective Color layer (while the selection stays unaltered)
/> Create a Hue/Sat.-layer, preferably in Saturation blend mode (+ mask from the selection). Confirm by OK.
/> Alt + click on the mask
/> set a Threshold with the respective tool (to protect objects/areas which are already quite saturated).
/> apply a Gaussian blur of some pixel width
/> tune saturation depending on needs

Now I’m sure that I’ve forgot or confused anything. If so, please let me know.

Peter

--
Logged

Chris_T

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 541
trying to understand saturation
« Reply #6 on: July 08, 2006, 08:08:27 am »

Quote
That’s how it is.

Finally, the Hue/Sat.-tool lets the data bump against the border of your RGB working space which is indicated by R or G or B= 0 or 255.  When this happens with too many pixels, posterization can occur.  Possible solutions:

/a)  To convert to a larger working space like ProPhoto RGB which offers more headroom.  That way, there’s of course a certain risk to overdo it; to create fancy saturated colors (or even imaginary colors) which lead to output-referred problems (monitor / print). So it can be wise to combine this with a Selection e.g. as suggested in point d).

/b)  Saturation Curves, e.g. such as offered by the Curvemeister plug-in.

/c)  To convert to Lab mode and to work with S-shaped curves on both color channels, a & b.  Probably great, but not really my field of competence.

/d)  Saturation Masks.  Just recently, I’ve learned & implemented the following technique:

/>  Create a Selective Color layer. For all R, Y, G, C, B, M color ranges set Black to -100%, respectively. For the White, Gray, Black ranges set Black to +100%, respectively (everything Absolute).
/> Strg + click on the RGB composite channel (to select everything visible)
/> Strg + Shift + I to Invert the selection
/> Delete the Selective Color layer (while the selection stays unaltered)
/> Create a Hue/Sat.-layer, preferably in Saturation blend mode (+ mask from the selection). Confirm by OK.
/> Alt + click on the mask
/> set a Threshold with the respective tool (to protect objects/areas which are already quite saturated).
/> apply a Gaussian blur of some pixel width
/> tune saturation depending on needs

Now I’m sure that I’ve forgot or confused anything. If so, please let me know.

Peter

--
[{POST_SNAPBACK}][/a]

This author has another way of creating a saturation mask, and has an action at his site. Have not tried either, and wonder how the masks differ from yours.

[a href=\"http://luminous-landscape.com/tutorials/restore-clipped.shtml]http://luminous-landscape.com/tutorials/re...e-clipped.shtml[/url]
Logged
Pages: [1]   Go Up