(De)Saturation Techniques: 1
Following on from my photoshop vibrancy challenge, I thought I would talk about some photoshop techniques I’ve been playing with for changing the saturation on photographs. All thumbnails are clickable.
Original shot
Apologies to my niece Megs: I took this photo without letting her pose for it with the result that she doesn’t have some fake grin and posture in it 
Technique 1: Colorize in Hue Mode
a simple but effective technique I picked up last summer:
Create a HSB adjustment layer, colorize, then use the ‘hue’ blending mode.
the greys are desaturated, and although the tone is monochrome, you retain some of the depth and range of the original.
Technique 2: Duotone in Hue Mode
The first image here is the tritone, included for reference.
Once you’ve created a duotone (or tritone in this case), you can copy it back as a layer to the original, again setting it to a Hue blending mode.
The duotone gives you an extra element of control over method 1 because it allows you to define the colour at different ranges (i.e. curves of shadows, midtones & highlights). However, you can clearly see that the resultant colours can’t be judged by those of the tritone you’re using: compare the colours here to the previous image in the set (the very tritone used to set the hues).
Technique 3: Using a Saturation-based Mask
Here’s the fun bit: thanks for my friend Marcus for the first part of the recipe.
We’ve created a channel based on the saturation of the original image and used it as a mask for a HSB desaturation of -90.
This is perhaps not the best sample image to show the effect (which loses a lot of effectiveness at small sizes), but if you look closely you can see that the center of the daisy is still a bright yellow.
‘Saturation Channel’ Recipe (Tom’s version of a Marcus original):
1. Duplicate the layer(s)
2. Hide the original layer(s)
3. add a layer of solid red, blend mode Hue
4. add a layer of solid red, blend mode Luminosity
5. Flatten visible (this should leave you with an image that goes from gray to red).
6. Use the channel mixer: I used red+150 constant-30 with the monochrome option checked to give a grayscale result
7. copy the layer and paste it into the quickmask to create a selection
8. Save the selection.
In this particular image I mucked about a bit further when using the selection as an image mask with brightness and contrast.
Technique 4: Posterizing the Saturation Mask
The Saturation-based Mask is a basis for a lot of experimentation: In this case, I posterized the mask on the HSB adjustment layer.
I think the result is remarkably subtle.
Thanks to Ian Tindale for the suggestion.
note: I originally posted all of this as a set of six pictures on flickr, but I decided it was wrong there: it cluttered up my photos, used up a valuable set (free accounts only get 3). Besides, I’d want to do a comparison like this, and it’s a discussion, not a portfolio.
So I’ve replacing the flickr stuff with a composite image and a link here.
May 11th, 2007 at 6:32 pm
Your niece doesn’t look anything like you!
I think I personally favour the duotone image
May 11th, 2007 at 7:56 pm
This might be cheating a little, but Paint Shop Pro 7 on Vista lets me reduce the saturation as seen here:
http://www.everythingeverything.co.uk/files/desat.png
Very simple, but very effective.
Your first two techniques make the yellow daisy look a bit too orange, which I don’t particularly like. I like the third one, although it’d be nice to see it on a slightly better photo.
Here’s what happens when I use -25 on your niece:
http://www.everythingeverything.co.uk/files/desat25.png
I think the amount you desaturate really depends on personal preference and the original photo.
May 11th, 2007 at 9:51 pm
Fab: No, but her mum does a bit
SB: It’s not just a case of desaturising it, but of selectively doing so. The grey bits get greyer faster than the not so grey bits. I’ve not finished playing about with it yet, but to get an idea of some of what I’m going for, try looking at the photos linked to in the last post.
The idea wasn’t simply to desarate, but to explore and compare the methods. This picture is a good one for skin tones, but I want to do something similar with a picture with much brighter colours. Watch this space.
May 11th, 2007 at 10:38 pm
The problem I see with your first couple of ways of “selectively desaturising” it is that the colours get changed, so that the nice yellow colour becomes a rather nasty and unnatural looking orange. Unless you specifically want that odd colour.
The third technique, based on Marcus’ settings, keeps the nice yellow colour; but the overall picture is fairly light, especially when you compare it to my quick and dirty “desaturate” approach that took a lot less time. But the desaturated image looks poor because it’s mostly greyscale, but with strange looking patches of mousey coloured hair and pinkish ears (which might be interesting if it weren’t for the fact we’re used to seeing kids with consistent colours on their face).
Your last image is quite nice, it has a washed out look, but comes across as warm and natural.
I’ve also been playing with the histogram functions, such as changing the luminosity based on a nice curve, fiddling with the contrast. But how you combine all these seems to depend on the original photo. For instance, if the original photo doesn’t have much contrast in the hair, you may need to fiddle with the settings in order to make the different strands and contours more visible (unless you intentionally want to make it consistenly bright/dark in order to emphasize other areas of the face/photo).
Talking of selective, it seems many people like to use the filters and then apply/blend it into selected parts of the original image. Useful if you decide you want to use different settings to make the clouds in the dark sky look more menacing and full of different greys, without making the face into a washed out or overly dark mess. Various filters applied across the entire image might not return the result you’re looking for.
Whenever I look at the JPEGs you’ve put online, the things that catch my attention the most are the artefacts and high compression, which then look even worse once you start to manipulate it with various filters. I saved my image as PNG, which probably makes it easier to spot the flaws in the original JPEG that allows you to clearly see square blocks around the corner of Meg’s left eye (you can see them in the original if you look closely). On your original “challenge” post I managed to play with various settings that revealed a lot of different shades of grey in the girl’s shirt, but it also highlighted the noise/compression, which made it look bad.
Sorry, that was much longer than I’d originally planned. The worst part is I prefer the original image, as it’s mostly pinky/browny/yellowy with a nice contrast on the pink lips and yellow daisy. The daisy has a natural vibrancy that makes it lovely to look at, you don’t really need filters to make it look “better”.
I think one weekend we should have a competition where we have half a dozen different photos (high resolution/low compression) and half a dozen different themes and you can mix and match the themes to the photos and do whatever you like, and then post the results online somewhere for everyone to see.
May 11th, 2007 at 11:05 pm
I probably prefer the last one, too.. If I weren’t doing it as a proof of concept, I would probably use that method less drastically and end up with something a little better than the original (perhaps). Really this wasn’t the best image to start this off with: I have a plan to take a better one when I come across the relevent subject matter.
Sorry, SB, I was a bit miserly saving those jpegs: the raws were a bit better, honest.
Filters that play with the entire image is what I’m after, here… sort of in the spirit of a darkroom technique rather than a digital graffiti cowboy. Not that there’s anything wrong with that, that’s what I wanted to play with when I got myself a digital camera: but I’ve begun to fall in love with the pictures that you never use a brush with. It’s just what I’m playing with right now.
I’m not quite sure what you mean by your competition, but I’d be glad to play along (and will provide any raw files for any images I’m asked for).
May 12th, 2007 at 10:22 pm
[…] Following on from part 1 […]
May 12th, 2007 at 11:45 pm
http://marcus.minotaur.cc/sundry/Effect2-1.jpg
That’s a combination of my selective desaturation, a duotone-alike thing, and the high-pass filter giving it that hard, war-torn look.
May 13th, 2007 at 12:14 am
Unfortunately I think the high-pass filter is mainly bringing out my overzealous jpegisation.
Nice tones, though.
May 13th, 2007 at 1:21 am
Yeah, when I first saw the saturation map it looked like it was made of huge pixels.
July 17th, 2007 at 3:00 pm
I just made use of this technique, or one similar to it, on a picture that was overexposed in daylight: I made a mask of the most saturate areas and applied that mask to an inverted copy of the image. The result is an image that’s only subtly altered to tone down those parts that were just too much.
July 17th, 2007 at 3:25 pm
linky linky!
July 17th, 2007 at 4:00 pm
It’s unfair to tease us like this! We want a link!