Options
LAB and B&W Conversions
rutt
Registered Users Posts: 6,511 Major grins
I started this thread as a place for a discussion that started as a hijack on a different forum.
If not now, when?
0
rutt
Registered Users Posts: 6,511 Major grins
Comments
1. Convert to LAB
2. Make the A and B channel flat/horizontal. (For pure b&w the horizontal goes through the middle (0 or 50%))
3. Play with the Lightness channel to add/reduce contrast.
4. Sometimes I use Unsharp mask on the Lightness channel to increase local contrast (large radius, low amount)
5. Sharpen final image.
Is there any advantage to use LAB over RGB or vice versa?
-- Anton.
When I hear the earth will melt into the sun,
in two billion years,
all I can think is:
"Will that be on a Monday?"
==========================
http://www.streetsofboston.com
http://blog.antonspaans.com
That question is like the first move in a really complex chess game. If you can, get Dan Margulis' Professional Photoshop and read the chapter Friend and Foe in Black and White. In short, the technique you outlined works, but loses the color information too early. For example, in my conversion of the woman, I used the blue channel for the hair and the green channel for most of the rest. Often the red information is very bad for faces but good for skys. Photoshop converts to LAB without knowing about faces and skys. But you do.
The conversion has resulted in a nice strong picture with deep blacks and pure whites, BUT technically it could be better. The policeman's shirt in the foreground has "plugged" (lost all shadow detail.) Look at the red and blue channels in RGB. I'll bet you can find some of that detail. Good B&W conversions exploit the detail in each channel. Blindly convert, either in LAB or just by converting to gray scale, and you lose the opportunity to do this.
Actually, a lot of the details in the shirt were left, but got plugged when i increased the dynamic range. I just tried to do the RGB channel mixing, but i never got a nice contrasty picture - unless when i increase the dynamic range again. And that removed the details from the shirt again.
Short of brightning/darkening of only selective parts of the picture (much more work in PS:D ), how would i go about that?
If i understand LAB correctly, is the L channel not the channel that contains all the detail, where A and B contain only the colors? The shirt is just blue-ish (no details here) and the L channel contains the shadow details.... So, removing the A and B channel should not matter much, should it?
When I hear the earth will melt into the sun,
in two billion years,
all I can think is:
"Will that be on a Monday?"
==========================
http://www.streetsofboston.com
http://blog.antonspaans.com
You're right let's move it to that new thread.
For now, i reply here with the attachment of the color version.
The color version (shot in RAW), has been resized and slightly sharpened.
When I hear the earth will melt into the sun,
in two billion years,
all I can think is:
"Will that be on a Monday?"
==========================
http://www.streetsofboston.com
http://blog.antonspaans.com
Photoshop determines the luminosity by a simple formula:
That's all it does. It just looks at each pixel in the color version and determines what shade of gray to use based on that formula. Conversion to LAB applies a gamma curve which lightens up the L channel a bit compared to conversion to gray scale, but the differences is slight.
Does this formula do the right thing? It depends on the image. Here is a contrived example:
Use your method of converting to LAB and flattening the A and B curves and you get this:
No amount of playing with the L curve is going to introduce any contrast between the stripes after this move.
On the other hand, I used the channel blender to make this B&W conversion:
See, I got to decide how to treat each color as it was converted to gray. And that's a decision I don't want to leave to some piece of software.
So what does this have to do with your image? Well, do you want this?
Or this?
This?
Or this?
The difference between these conversions is how I blended the red green and blue channels. The first version is my favorite becaue I was trying to show detail in every part of the image and I like that shade for the helmets. But the real point is the decision should be yours.
If you want to play with this a little, download: http://gate.chezrutt.com:8030/rutt/cops.psd.zip and play around with the opacity and blending options of the layers I made. You won't have any problem producing the versions I made as well as an infinite number of others.
There is a huge amount of lore on dgrin and beyond on this topic. The very best thing you can do to understand how to make great B&W conversions is to buy Dan Margulis' Professional Photoshop and read the chapter Friend and Foe in Black and White. Nothing else will take you as far, and in the long run nothing else is easier because nothing else addresses the root issues as directly or as deeply.
Short of that, dig through the stuff on dgrin. Beware that there is at least as much chaff as wheat here. This is as good a place to start as any: http://www.dgrin.com/showthread.php?t=14266
I don't know how to say this differently, but here it goes:
Thank you very much!!
I'll give the cops.psd a go and see what i can do.
I remember i have a photoshop action somewhere on my PC that does the same. It makes a picture B&W in one action and after the action is done, you can set/modify the percentages of the R, G and B channels and modify or remove some other effects (grain, tint, highlight).
About the R+G+B percentages: Must they always add up to 100% to get a 'proper' b&w? I guess the answer is 'yes' if one wants to preserve the luminosity...
When I hear the earth will melt into the sun,
in two billion years,
all I can think is:
"Will that be on a Monday?"
==========================
http://www.streetsofboston.com
http://blog.antonspaans.com
Actually the answer is yes by definition. It's 100% of the total no matter what that is.
I always find the channel mixer in RGB mode the best means of producing an image. Usually i would look at each channel by itself to see which would produce the deepest blacks. From there I have an idea which channel amount i should use.
But i was told that usually using mostly a blue channel in some situations produce increased detail.
Also, using the HIRALOAM method in lab before moving into rgb is helpful for producing greater contrast of values, as mentioned by a previous poster.
Chris
Detroit Wedding Photography Blog
Canon 10D | 20D | 5D
When you blend the channels into a gray scale image, what you should be doing is choosing the relative shades of gray for the various colors in your image. The policemen's helmets were blue in the original shot. Should they be lighter or darker than the faces in the B&W? The faces have detail in the orignial which is probably nonexistent in the red channel but very visible in green and blue. You need to pick a blend that maps that detail into different shades of gray so it isn't lost. After you do this blending, you can always used curves or levels to make to get the blacks and whites right. The part that only you can do and which has to be done by channel blending is resolving the ambiguity of mapping the colors into gray scale.