Ben King's Comment on DV/HD color space?

I am just searching for enlightenment, on the Monitor Flicker query Ben King commented on HD color space being different to that of SD and I don't care how thick I sound making this request but if anyone, Ben you to, can expand and define the concept of color space for me I would love it. Is it a generic or scientific term for the color of various video formats. Yes, I have a rough idea but the science of video is something I am still learning about even though I have edited for seven years. 'Pull down' is another term I should be up on but am not. Cheers Phil UK

Haha. you're from the UK. Pull down is almost alien to us PAL users. When converting between film to PAL, the film is simply sped up around 4%. In NTSC, however, a pulldown is used. And yea, it's something that would confuse us PAL editors.

this seems like a good page for that if you want to know more. [www.paradiso-design.net]

There's a lot of math involved in understanding Chroma subsampling. What i understand from it is that it's ratio of luma to chroma compression, as humans are more sensitive to luma movements than to chroma movements. As video requires compression, one way to achieve that compression is to reduce on the color space by sampling at lesser points than with luma samples.

RGB codecs are always 4:4:4 codecs (eg. Animation, microcosm, etc..), then there are 4:2:2 codecs and formats that are frequently in use (Uncompressed 4:2:2, prores hq, digital beta, dvcproHD, dvcpro 50), then there are 4:1:1 codecs that compromise further on the color space (dv ntsc, dvcpro 25), and finally, the 4:2:0 codecs (dv pal, HDV, mpg).

4:2:0 is a little harder for me to understand, but it just doesn't look as good as 4:1:1, which is more consistent both vertically and horizontally.

Corrections, anyone?

4:2:0 has two variants - one for interlace, one for progressive. The progressive one just stores chroma in 2x2 pixel chunks. The interlace one does also, but on a 2x2 chunk in a field, rather than in a frame, which must be decoded with care.

Colour spaces are defined by standards and include the matrix that does XYZ to RGB transform and the curve that defines it's gamma curve. For HD the standard is REC709. For SD it's REC601.

Best guide for all these technical things is Poynton's book. www.poynton.com

Graeme

[www.nattress.com] - Plugins for FCP-X

Haha. I just get lost whenever i start seeing the numbers... Never was a good math student.

Graeme has a great article to get started too! [www.nattress.com]

Gamma Curve...this is another term I need to understand...I use the Natress filters and gamma curve is part of the process and I stay away as I have not got a clue how it effects what I'm doing.
Thanks guys, Strypes whoever you are, we speak often. I'll check the link. Are you UK based? I've chatted with Ben King a couple of times. I am always keen to speak to UK based FCP editors as Avid still plays a dominant role although in smaller production houses FCP is raiding the Bastille. I'm much more comfortable on FCP than I was on an Avid and the filters and effects give FCP a great deal of edge over other platforms. Avid Symphony is fantastic but what an expense.

Wish I was. I'm from Singapore, Asia. And am doing my rounds in the smaller houses and learning everything I can so that I can soon move on out of here and get decent gigs on projects that don't make me barf so much...


> although in smaller production houses FCP is raiding the Bastille.

Final Cut has crazy versatility for the price! From what i remember, the speed ramps are still inferior to the fluid motion in the Avid Adrenaline. Then again, you want a one stop solution to post production and don't want to shell out incredible amounts of cash on Pro Tools, Avid, Smoke, AE, a top end encoding and authoring software? Final Cut is the way to go. Not that i'm a die-hard at selling it. Motion is no AE, Soundtrack Pro is no Pro Tools, you'll still need a good capture card to run that stuff into a deck, but hell, you get Logic plug-ins and Avid guys without the extended range of 3rd party app support can only watch on and look. And as an NLE, it lacks the Avid on some features (script logging, but i never had a use for it, not without a capable assistant), but overall, the workflow is more streamlined with its 3rd party apps. For some additional features, i'd wish Final Cut deals better with interlacing, and that soundtrack pro can read omf files, a blu-ray authoring software as well as an in-built burner. Downside is that with all this versatility, we're now supposed to know everything!

Btw, Gamma is similar to "mids" on the 3-way color corrector. It has little to do with highlights or blacks. The different curves can be likened, in terms of luma, to frequency response, in terms of pitch, in the world of audio.

As a side note... I've been trying to figure out how to get the lift and stamp function (http://www.apple.com/logicstudio/soundtrackpro/#dialogue) to work in Soundtrack Pro. Anyone has any ideas how to get the audio filters to get onto that palette?

Phil-

Graeme's definition of Recs 709 and 601 is correct. As is his reference to the work of Charles Poynton. To really understand this stuff I would recommend a college level understanding of calculus.

For my own purposes I have had to lower the bar a bit. I can just about balance my checkbook. With the help of Quicken and patient bankers.

If you read Poynton you will discover that a lot of this stuff is based on the early use of tube based television equipment. The first thing to understand is the gamma curve.

As we are all aware there is a relationship between voltage and brightness in a television set (or CRT, Cathode Ray Tube). This relationship however is not linear. If we raise the voltage in a linear way from 0 to 1 volt, the resulting change in brightness at the face of the CRT will present as a curve. We call this the gamma curve.

To fix this we could put a control at each TV to "flatten" the curve. Early TV folks figured the cheaper way was to put the control at the camera, and make the adjustment before transmission.

Then we got to color. Because we apply the gamma correction to each of the RGB channels prior to encoding into color signals we screw up the idea of having equal amounts of Red Green and Blue to produce blacks and whites. That is the matrix formula that is mentioned in this thread.

The spec for SD is defined by SMPTE REC 601. The math is:
Y' = 0.299 R' + 0.587 G' + 0.114 B'.

When they defined HD for some reason the felt the need to play with the matrix to produce:
Y' = 0.2126 R' + 0.7152 G' + 0.0722 B'

This means that if you up or down-convert between SD and HD you need to be sure that you apply an appropriate color transform. In my (limited) experience this is handled by the software or hardware doing the conversion.

BTW I found the matrices on Wikipedia on this page:
[en.wikipedia.org])

I have also seem some information on Wikipedia about pull down, but as you are lucky enough to be British it doesn't really change your workflow.

-Vance

Hey Phil

Hopefully this has pointed you in the right direction. Whilst its not a necessity to understand the serious technical background to colourspace encoding or conversion, it is something to be aware of and to know how to correct or allow for when converting footage or exporting for different display devices and formats.

Apart from the links the guys gave you check Wikipedia or search google - there are literally thousands of articles out there on video technology that you can read.



For instant answers to more than one hundred common FCP questions, check out the LAFCPUG FAQ Wiki here : [www.lafcpug.org]

This is great..I'm all for it. Video is a science and it is developing in wild and wonderful ways and we'll be there to see it. Calculus? I would have a go but a = b = c messed me up when it came to maths. I do have a friend who is a profound mathmatics genius teacher who explained the quadratic equation to me using satelite beam and the angle at which it hits the dish as a model.

Color space is the 3-dimensional realm of perceivable colors. In around 1800, Thomas Young figured out that there were too many colors for there to be a separate nerve for each color at each spot on the retina. He surmised that a color is the consequence of just three signals (as could be carried on three nerves). In around 1860, Maxwell demonstrated that every color could be matched by a sum of three arbitrary primary lights, if you allow for negative amounts in the sum. Choosing highly saturated R, G, and B primary lights lets you match most colors without need for negatives. That's about where we are now.
Knowing that color space is 3-dimensional doesn't mean we know what the dimensions of color space are. RGB space and all the CIE color spaces are unsatisfactory because they distort the perceived color space. In around 1940 MacAdam showed that the perceived color space -- that is, colors arranged according to their perceptual nearness -- is non-Euclidean and much messier than any of the models.
Anyhow, that's what color scientists mean by "color space". It doesn't help with Ben King's comment.

Dennis Couzin
Berlin, Germany

Seeing as this thread is expanding into a science lesson...

You perceive colour and luminance via separate photo receptors in the retina (called Cones & Rods respectively) and the two are processed in different parts of the brain too!

We have three different types of cone (Long Middle & Short Wavelength Cones) but only one type of rod (he's a lonely guy) which is why we can't distinguish colours (lucky or educated guesses don't count) in very low light.

A must read for everyone in the visual arts is:

Vision and Art: The Biology of Seeing

By Margaret Livingstone

ISBN: 0-8109-0406-3


Also a good thing to have lying around the suite is a load of eye tests for colourblindness in case "that client" comes back and tells you your $8000 Broadcast monitor is show the red of his logo as pink...

Seriously I have stories about "the client is always right"... until he went to the optician he was!



For instant answers to more than one hundred common FCP questions, check out the LAFCPUG FAQ Wiki here : [www.lafcpug.org]

The rods of the retina have nothing to do with how we see video since (1) the rods are absent from the cherryspot (fovea) of the retina where we see high detail, and (2) the rods are bleached out at the luminance levels of videos.
It is true that we see luminance by means of the rods, but only in twilight and night.
We see luminance by means of the cones by day and when watching video. Thus to understand how we see video it is enough to understand how cone vision works. It is because there are three types of cone that color space is 3-dimensional. An aspect of color space is luminance, for example, yellow and brown are different colors even though they may have the same chrominance: they differ in luminance. The rod system provides the luminance information for day vision and when watching video. That luminance signal is approximately the (weighted) sum of the signals from the medium and long wavelength cones.

Dennis Couzin
Berlin, Germany