UNTITLED - by Zach, Mac, and Jack

Seems like it pushes or squeezes the pixels into newer, smaller squares creating a new color space arrangement

(What the hell am I talking about?)
 
FatBird19 said:
yes.

I've also noticed that shooting in 1080x1280 on the HVX (given that the chips are only 540x960) also comes out to about 4:4:4 color resolution.
Click to expand...
4:4:2.6 - you're losing some red

As you can see:
YPbPr (analog version of YCbCr) from R'G'B'
====================================================
Y' = Kr * R' + (1 - Kr - Kb) * G' + Kb * B'
Pb = 0.5 * (B' - Y') / (1 - Kb)
Pr = 0.5 * (R' - Y') / (1 - Kr)
....................................................
R', G', B' in [0; 1]
Y' in [0; 1]
Pb in [-0.5; 0.5]
Pr in [-0.5; 0.5]
Y' = Kr * R' + (1 − Kr − Kb) * G' + Kb * B'
 
The reasoning behind the downconverting giving you higher color sampling is that the downconverted image will be losing half its resolution in each direction, but the color sampling is already at half resolution in each direction -- or, there's enough color information to make for a pixel-accurate map (or, nearly pixel-accurate; probably close enough in any case).

In other words, you have, say, 1440 x 960 in luminance, and you want to downconvert that to 720 x 480. If your 1440x960 image was processed at 4:2:0, it would have a subsampled color resolution of 720 x 480 already. So if you could scale the luminance and leave the chrominance alone, you could potentially be at 4:4:4 (every pixel having its own color).

That's the theory, anyway. The reality is that you're going to do to your resolution the same thing that chroma sampling does to your chroma. 4 pixels will become one. And the color of that pixel will be the average of all four pixels. Chroma sampling has already taken care of averaging the color together, so in the actual downconvert process you'll have four same-color pixels being converted into one.

So yeah, you'll get a pretty good approximation of 4:4:4 SD when you're done with the downconvert (but when going to DVD it's going to get resampled at 4:2:0 anyway, throwing away 75% of that chroma you were trying to retain... sigh.)

But then it gets a tad more complex if you're talking about 1280x1080 or 1440x1080 or 1280x720, none of which have an exact scaled 2:1 relationship to 720x480 or 720x576, so there's going to be some fudging going on somewhere anyway. And it gets worse with interlaced 4:2:0 chroma, so if you're going from 1440x1080i HDV down to 720x480i DV, it's not going to be as clean as it would from progressive or 4:2:2 interlaced.
 
Even if that were the case, 2 colors right next to each other are still not sampled as two separate colors with 4:2:2 color space.

They will be sampled as shades of the same color range that the camera thinks most closely approximates the actual colors. So, even if in the downconvert process things are getting re-sampled/squeezed whatever, at 4:4:4, they are still getting resampled from the camera's "approximation" of what the original color was. Does that make sense?

Let's use black and white right next to each other. Now that (only an example) may get sampled as a dark grey by one pixel, and a real light grey by the other, in the two pixel group that is sampling that portion of the screen.
So if you are getting this 4:4:4 downconvert, what would that mean? The downconvert process is not going to be able to magically know what those original two colors were. If it could it would have to be a really aware piece of software.

I've gotta stop, cause we are supposed to be talking about your Drama Fest entry MacGregor, which I am totally looking forward to, and digging your poster (as I mentioned earlier). Plus I haven't the technical know-how or knowledge to know or explain these things for certain. It just doesn't seem possible to make 8 dollars out of 5 dollars without doing some kind of investing (to use a bad analogy)

Later,
Jason
 
dangit barry, why'd you have to go and make my head explode.

mcgregor / zacharias .. im very much looking forward to your film. :)
 
Just BTW - Even High end telecines at the moment mostly output to 4.2.2 (including the spirit datacine - at least in Europe). 4.4.4 in the broadcast industry (and by that, I mean finishing on TV) is generally done at 4.2.2 (4.4.4 is rare) - even for jobs with heavy post/ greenscreen work.

Having said all that 4.4.4 is a dream to key with. Hair etc generally is less of a problem.

or you could just cast bald actors and shoot 4.2.0
 
macgregor said:
Not at all.
I would obviously prefer to use a HVX, but i must find one in Utah!
Click to expand...

Up until three months ago I lived in Utah, Macgregor. If you need help tracking down a HVX or DVX to use there, let me know. And if you need any crew, I would be down for making a special trip back there to help you out. :smile:
 
Hi Mojo! Did you get my PM?

We urgently need a Jeep Wrangler, a new Audi TT, a Ferrari F430 or 599 GTB and two actors for the filming:

He should look like Philip Seymour Hoffman, in his 35-45s, shy and good person feeling.

She should be similar to Natalie Portman, very femenine, atractive but not exhuberant. Someone in which you would fall in love.

Whiout all this the film is not possible i am afraid.
 
macgregor said:
Hi Mojo! Did you get my PM?

We urgently need a Jeep Wrangler, a new Audi TT, a Ferrari F430 or 599 GTB and two actors for the filming:

He should look like Philip Seymour Hoffman, in his 35-45s, shy and good person feeling.

She should be similar to Natalie Portman, very femenine, atractive but not exhuberant. Someone in which you would fall in love.

Whiout all this the film is not possible i am afraid.
Click to expand...


Just sent you a PM, Mac. I'll keep a lookout for the vehicles too.

Half your films have nude girls in them. That might be harder to manage with Utah actresses. You might have to import a hot girl from outside the state :evil:
 
You must log in or register to reply here.
Back
Top