View Full Version : CMYK
GenJerDan
01-05-2005, 05:43 AM
Is there an fx filter (or a script) that will let me split a vid into its CMYK components or drop everything but the C, M, Y, or K?
Dan
XCheck
01-05-2005, 07:24 AM
Not as far as I know. You would have to have the source in CMYK color space - I don't think this can be done 'on the fly' like converting from RGB to HSL and back. CMYK is a four-channel color space, so the transformations are more complex - it's probably a performance issue rather than anything else.
Try to export your footage to PNG or JPG image sequence, and do it in a batch in Photoshop (don't forget to convert back to RGB, otherwise Vegas may not be able to import the image sequence back).
J_Barnes
01-05-2005, 08:42 AM
No. Video doesn't have CMYK components, so there's no process out there to do what you're saying. You can do something with RGB, but CMYK would require a conversion to CMYK and ultimately back to RGB as that's the only way you can display video in the real world (unless you're looking to make a flip book!)
GenJerDan
01-05-2005, 08:46 AM
Ah well.
Just for messing around with, anyway.
I grabbed some frames as PNGs, CMYKed them in PaintShop, then un-CMYKed them. I liked the way the colors came out. (It didn't just go back the way it was)
Actually, what I stared out wanting was any way to separate the "layers" (yeah, I know, video doesn't have layers as such) then "re-sandwich" them like I was doing 3-strip film.
Like I said, just messing around. :)
Dan
XCheck
01-05-2005, 09:22 AM
AHH, you CAN do that! As soon as you used the magic word LAYERS, I got an idea and tried it right away... I may not be the exact CMYK, but close enough...
Take a track, and duplicate it four times. In three of the tracks, use the color curve filter. Select R, G, or B channel in each of the tracks respectively, and pull that channel's upper node along the Y-axis all the way down. You will first need to set the tangents to be horizontal on each point, and then uncheck the 'lock tangents' to be able to do that.
So, when you pull down the channel, you will get its complementary color:
R->C
G->M
B->Y.
For the K channel, use HSL adjustment first, set saturation to zero. Then apply Color curves to set the K-channel curve the way you want.
Mind you, this will not be TRUE CMYK, because, if you take a medium gray (a point in the middle of the curve), you would end up with 50%C, 50%M, 50%Y, and 50%K, while you should probably have something more like half of it, but once you've separated the channels, you should be able to adjust the individual curves the way you want...
Hope this helps.
Jerry
GenJerDan
01-05-2005, 11:53 AM
Thanks.
I tried everything but the Color Curve plugin. ::)
Ooooooo....gonna have SO much fun this weekend.
Dan
XCheck
01-05-2005, 11:56 AM
Let me (us) know how it worked for you!
David Jimerson
01-05-2005, 12:12 PM
Well, I've been schooled. You photography people . . . ::)
:D
XCheck
01-05-2005, 01:09 PM
There is this book "Adobe Photoshop for Photographers"? (I am [mo]mentally memory challenged, so I don't remember if that's the one) that explains everything so well in terms of CMYK that I don't want to color-correct in any other space. For example, it shows you how to get a perfect skin tone in CMYK using just numbers (forget about calibrating your monitor, for colorblind people like myself it's no use :D), do selective channel sharpening to get the eyes and lips stand out, and so much more. Channel separation and blending - something I never understood from Photoshop manuals.
Anyway - CMYK rocks, it's much easier to deal with than RGB. It is more natural because when we grow up we learn to deal with subtractive color mixing every time we use a color pencil.
I'd say that RGB is scientific and technical - CMY(K) appeals more to the artists. I say (K) because there is a school of thought that black doesn't exist in nature (except, of course, black holes and black bodies at 0 Kelvin temperature), so why bother mixing it in - you should always be able to mix a color with the basic colors only. That may be true in painting, though. In color photography and video it's another matter - I guess due to the limited color space to start with.
I hope that Dan becomes the pioneer in the use of CMY(K) color correction in Vegas!
J_Barnes
01-05-2005, 01:20 PM
The problem with CMYk for video is that there is no CMYk in light (projection, monitor, television)...it's impossible. It only exists in terms of ink on media...every non-ink version of CMYk is really just an electronic emulation displayed in RGB, not a true CMYk.
Anyone who works in print should learn more about CMYk, trapping, etc. But it has little use in video or projection.
Actually, the only thing that comes close to CMYk is bleech-bypass film projection (which is actually RGBk)...but still that does not truly exist in an RGB-only environment like computer monitors, projectors and television screens.
XCheck
01-05-2005, 01:44 PM
Absolutely agreed. I guess it's the 'color correction' aspect of it - at least for me it's easier to understand how to adjust CMY values than RGB. As far as the actual projection goes, you are correct - there is no CMYK in video because it's all additive (painting with light, not ink).
On the other hand, that book I mention basically says R=C, G=M, and B=Y, so if your picture looks too yellow you just add blue. I just haven't figured out what to do when it's too red, without changing luminosity as well. I tried HSL but not much luck there, either - the whole color space rotates, so I may get some colors right but others get way wacky.
I wish I could post some grabs to show what I am talking about, but I have this outside shot just after dusk, with street lights (I believe they were sodium vapor), and my white balance was still set at 5.6K (which had worked quite well until about fifteen minutes before dusk - I guess the mix of street light and some sky light was allright, but once the skylight became too dim it went all red (almost magenta-ish).
In CMY, I'd just adjust C,M,Y on people's faces until I'd get the right values, and everything else would neatly fall in place.
That would be in photoshop, though - I am still learning how to do this in Vegas. It would be very useful if there was something like the sampler tool in PS, where I could see the values of a particular pixel as I adjust the curves or whatever.
Jerry, why not just go in and adjust the curve for the red channel? Am I missing something?
J_Barnes
01-05-2005, 02:15 PM
Jerry, I'm pretty sure I've read the same book, and indeed it was enlightening to learn that the best color correction for print can employ three entirely different color spaces. Sharpening in LAB, and correcting in CMYk can both give you results you can't get at all in RGB.
Being an amateur printer myself, I don't dabble in CMYK. I can't afford a RIP for my printer, so I'd be doing three conversions before output (RGB>CMYk>RGB>Printer's CMYk).
Ultimately though, you're right. CMYk is the most logical, RGB the most reliable (if unweildy), and LAB is the most abstract.
XCheck
01-05-2005, 02:43 PM
Jerry, why not just go in and adjust the curve for the red channel? *Am I missing something?Tried that - it gets too dark. I also need to adjust the blue channel to remove some of that magenta, and people's faces turned green. Then I tried increasing luminosity with HSL - that sort of worked, but a red street car in the scene now looks really unnatural color. I guess that's real problem - how do you color correct something where the color you want to change is quite legitimate in a portion of the scene...
I won't have access to my files for the next week or so, but I'll try to post some grabs when I get to it, to see if anyone has an idea...
Dan, how is the CMY experiment working?
Hmm interesting sounding problem! Looking forward to seeing the grabs when you get them up!
GenJerDan
01-05-2005, 04:07 PM
Dan, how is the CMY experiment working?
Hey...it's 1 a.m. here. I'm sleeping.
Um...er...well, I will be, soon.
:)
Dan
XCheck
01-05-2005, 04:35 PM
So? That's when I begin to pay attention. ;D
Good night, sleep tight. Dream in CMYK color. :)
XCheck
01-06-2005, 02:10 PM
Dan,
actually, with the techniques I mentioned using color curves, you will get negatives of the individual color. You will probably want to pull up the lower-left points to get the right colors.
I am going to try it myself.
GenJerDan
01-06-2005, 03:52 PM
'k
I'll probably shoot something tomorrow to give it a try, too.
Don't know what...but I'll think of something. :)
Dan
XCheck
01-06-2005, 03:59 PM
I thought you must have had a lot of footage of that naked woman you were going to shoot last month. ;D
Or did your girlfriend swipe the tapes with a magnet? :o
GenJerDan
01-07-2005, 05:50 AM
Girlfriend is still 1000 miles away. :(
I wanted to go outside and get some nice nature background and such for testing...but I keep forgetting this is Germany and will be grey and rainy until April sometime.
I'll end up shooting me in the apartment being silly.
Dan
GenJerDan
01-08-2005, 01:53 AM
The experiment was a dismal failure.
Ah, well.
Dan
David Jimerson
01-08-2005, 11:15 AM
What's that GIF in your avatar, Dan?
GenJerDan
01-08-2005, 11:33 AM
One of the dangers of being in a video web chat room is that other people can see what you are doing.
Another danger is folks like me, who write programs that capture the vid for posterity. (Although, in this case, "for posterior" would also work...chats are generally 4:3, not 1:1, so there's more to the frame...)
Dan
David Jimerson
01-08-2005, 11:42 AM
So that's some girl who was on webcam in a chatroom?
GenJerDan
01-08-2005, 12:38 PM
There's just something about a pretty girl picking her nose that tickles me. ;D
But, yeah, too mean...
Dan
David Jimerson
01-08-2005, 12:54 PM
Didn't notice that was what she was doing. Was distracted by the hotness.
GenJerDan
01-08-2005, 02:45 PM
Didn't notice that was what she was doing. *Was distracted by the hotness.
;D Maybe I've just been chatting with her too long...the hotness wore off.
Dan