4K vs. 4K HD

[Philip Aromando]

4K vs. 4k HD? Very new to Red and embarking on a feature length documentary with the camera. Based on what I have read, 4K HD sounds like it's ideal for anything that's primarily going to be broadcast or screened in 1080 HD and 4K is more suited for theatrical releases. Is that a good rule of thumb? Or is it, most likely, more complicated than that?

 

[Gunleik Groven]

You're right

G

 

[Steve Gibby]

4kHD is mathematically 4 x HD in pixel dimensions, and natively in a 16x9 aspect ratio, thus it scales and resamples down to HD quickly and looks real good. I use 4KHD for basically all projects destined for HD delivery, plus the raw footage will then have a long useful period and multiple applications as the world eventually migrates to 4kTV, 4k monitors, and 4k projection. Well shot 4KHD will also upscale to 4k real well.

 

[Jacek Zakowicz]

I have a question:
If 4KHD debayers at 80% to about 3+K resolved does it mean that a downsampled footage will resolve half of that?
How does the downsampling work? Does it just bin pixels when going half?
Should we use footage that debayers to 4KHD so we get true full HD when we bin (downconvert ) in post?
I just don't know that's why I'm asking... It's been kind of bugging me....

 

[Frank Cueto]

Jacek, when you downconvert its important that you use the correct filtering. Pixel binning would be similar to using a "Near" filter. There are other options way more advanced that will actually take into account ALL the pixels provided to make the best downconvertion from all of them. Usually giving you naturally sharp images (no edge enhancement) and less noise due to the averaging of more than one pixels (oversampling if u may).

 

[Jaakko Rinne]

If you're shooting at a resolution that resolves up to FullHD when debayered (i.e. 2.5K and higher) your footage should resolve to FullHD when downscaled to FullHD. YMMV depending on the the downscale algorithm and other details, but you won't be off by much.

 

[Blair S. Paulsen]

Frank brings up an often glossed over issue which is choosing the best downsampllng filter for your material/deliverable.

I usually start with Lanczos or Mitchell, then try something else if I'm unhappy. Would love to know more about how the various options operate, if there are any significant speed issues and so on.

Cheers- #19

 

[Jacek Zakowicz]

Jacek, when you downconvert its important that you use the correct filtering. Pixel binning would be similar to using a "Near" filter. There are other options way more advanced that will actually take into account ALL the pixels provided to make the best downconvertion from all of them. Usually giving you naturally sharp images (no edge enhancement) and less noise due to the averaging of more than one pixels (oversampling if u may).

So if you are not binning what's the downconvert advantage for the 4KHD?
Seems that the higher native resolution the better. And the angle of view for wide shots is better with higher res....

 

[Gunleik Groven]

So if you are not binning what's the downconvert advantage for the 4KHD?
Seems that the higher native resolution the better. And the angle of view for wide shots is better with higher res....

Hey Jacek, you're of course right, too!

Bigger is better, but after a while (we and Jim may discuss exacctly where) the practical effect is diminishing....

4KHD has been a perfectly matched "good enough" aqusition format for most of the 1080 series I have been working with. Enough headroom for some reframing without loosing resolution. Filesizes compareable to 1080 prores 4444/Alexa, quick turnarounds and not to much computing overhead. One Rocket per station gives reasonable results as to speed.

Does that argument sit with you?

Cheers!

G

 

[Jacek Zakowicz]

Of course I understand the compromise here. If 4KHD gives faster downconvert via simple binning and the results are satisfactory or even visually identical to more complicated/time consuming algorithms then why not?
Have you tried both ?
Any samples of the same (reframed for 4KHD) frames to look at? Downconvert times? Anyone?

 

[Rob Ruffo]

4K 16:9 gives you a wider view on your lens, which will look better, and make whatever lens appear sharper - it is effectively giving you a larger sensor look.

Side by side, 4K 16:9 looks better than 4K HD, always, even in 1080p. Just try it on a good real 1080p 1:1 monitor - you will see.

 

[Gunleik Groven]

Of course I understand the compromise here. If 4KHD gives faster downconvert via simple binning and the results are satisfactory or even visually identical to more complicated/time consuming algorithms then why not?
Have you tried both ?
Any samples of the same (obivously a bit cropped in 4KHD) frames to look at? Downconvert times? Anyone?

But you're right...

We do not bin. Still going 4khd ---> 1080 is a quicker downsample than 5k ---> crop to 16:9 --> Downconvert

OTOH, when we need that extra detail, we frequently shot 4k 16:9 and now 5k FF on the Epic and diwnsample til 1080

We do NOT do binning, even though a few of the turnaround arguments suggest that.

But 4kHD is a nice match for turnaround/filesize/flexibility. Much better tech result than 3k, much more flexible tan 5k. Are we splitting hairs here?

G

 

[Jacek Zakowicz]

OK That's good to know. Are we splitting hair? In your experience? You've probably done and seen more tests than anybody here outside of the Red crew.....

 

[Steve Gibby]

If 4KHD debayers at 80% to about 3+K resolved...

Graeme Nattress and Stuart English have each said that the currently a full debayer of r3d files ends up at around 85%. I believe the 80% debayer figure was quoted in 2007 through 2009 with Red One files using Red Alert and the proto versions of Red Cine.

Well shot 4kHD, debayered, given a one-light, and resampled down to 1080p looks extremely good...

 

[Steve Gibby]

4K 16:9 gives you a wider view on your lens, which will look better, and make whatever lens appear sharper - it is effectively giving you a larger sensor look.

Side by side, 4K 16:9 looks better than 4K HD, always, even in 1080p. Just try it on a good real 1080p 1:1 monitor - you will see.

Yes, there's a slightly wider field of view with 4k 16x9 - but my friend Michael Cioni of Plaster City, who is widely respected in the r3d file editing community, feels there are many advantages to shooting 4KHD over 4k 16x9 for HDTV, BluRay, and online deliverables. He also feels that the perceptible sharpness of 4kHD is actually sharper than 4k 16x9. Here is his opinion on 4KHD vs 4k 16x9:

"At PlasterCITY, we have experienced 2 major advantages for shooting 4KHD over 4K16:9 when you are planning for a 1080p delivery:
First, the transcode times decrease by a significant margin. Depending on the speed of your machine, debayering tool and/or network capacity, the ability to process dailies can be improved by up to 35%.
Second, the perceptible sharpness of 4KHD is actually sharper than 4K 16:9 (again, when processing out for 1080p delivery).
The simple reason for this is the even divisibility of the originator raster size by 1920x1080. When debayering & scaling of 4K16:9 media, partial pixels appear "softer" than when the divisibility is whole. Eliminating the 46.875% scale (1920 divided by 4096) can eliminate a small amount of perceptible softening resulting from the leftover fractions.
The result, 4KHD is the best all-around shooting mode for television HD, BluRay & online delivery. The biggest trade-off is partial loss of the censor when shooting."

Michael Cione, Plaster City
Red User
1/22/09

 

[Jacek Zakowicz]

Yes, a slightly wider field of view with 4k 16x9, but Michael Cioni of Plaster City lists many advantages to shooting 4KHD for 1080 HD, BluRay, and online delivery. He also asserts that the perceptible sharpness of 4kHD is actually sharper than 4k 16x9. Here is his opinion on 4KHD vs 4k 16x9:

"At PlasterCITY, we have experienced 2 major advantages for shooting 4KHD over 4K16:9 when you are planning for a 1080p delivery:
First, the transcode times decrease by a significant margin. Depending on the speed of your machine, debayering tool and/or network capacity, the ability to process dailies can be improved by up to 35%.
Second, the perceptible sharpness of 4KHD is actually sharper than 4K 16:9 (again, when processing out for 1080p delivery).
The simple reason for this is the even divisibility of the originator raster size by 1920x1080. When debayering & scaling of 4K16:9 media, partial pixels appear "softer" than when the divisibility is whole. Eliminating the 46.875% scale (1920 divided by 4096) can eliminate a small amount of perceptible softening resulting from the leftover fractions.
The result, 4KHD is the best all-around shooting mode for television HD, BluRay & online delivery. The biggest trade-off is partial loss of the censor when shooting."

Michael Cione, Plaster City
Red User
1/22/09

That's interesting, "divisibilty by originator raster" sounds a lot like binning to me. I completely agree with "partial" or " out of phase" pixels looking softer however.
Also that was back in 2009 and it looks like today for whatever reason 4kHD is created by cropping in post (not as a shooting format in camera) after some reframing/stabilization work. Also "divisibilty by originator raster" is not the preferred downconversion type and other more sophisticated downconversion filters are used per posts above.
I'm confused here Gunleik, ?Gunleik??

 

[Steve Gibby]

Jacek,

When Michael made that post in January 2009 4KHD was an in-camera format choice with Red One. 4KHD is one of the in-camera choices for shooting format with Epic - and as I remember it has been for nearly a year now, since mid-2011. That said, 4kHD can also be created by cropping down in post from higher resolutions.

I'll let Gunleik answer your other questions...he has a lot more hands-on post production knowledge than I do.

 

[Frank Cueto]

That's interesting, "divisibilty by originator raster" sounds a lot like binning to me. I completely agree with "partial" or " out of phase" pixels looking softer however.
Also that was back in 2009 and it looks like today for whatever reason 4kHD is created by cropping in post (not as a shooting format in camera) after some reframing/stabilization work. Also "divisibilty by originator raster" is not the preferred downconversion type and other more sophisticated downconversion filters are used per posts above.
I'm confused here Gunleik, ?Gunleik??

When you bin, you are throwing away info. For example, if you where binning 4khd to hd, out of a group of four pixels, only one would be brought over IDENTICAL, to the hd downconverted image.. In the case of a proper downsample, all four pixel would be looked upon or analyzed depending on the filer to give the desired AVERAGE of those pixels (filter selection plays a HUGE ROLE here). Basically doing the equivalent of a supersamplig algorithm. Hope this is clearer than my previous post.

 

[Frank Cueto]

Oh, and on the argument of speed/sharpness. Down converting from 4k HD to hd is cleaner due to the lack of fractional pixels. The speed comes from the fact that the math becomes way simpler to process due to integer nature of the convertion too (god I hope I am right

 

[Jacek Zakowicz]

Thank you Frank
I actually thought that binning was averaging values from all 4 pixels always.
OK it makes sense now.
How does the debayer play into this? Before or after down conversion- which is better?
Mathematically is should not make a difference...