3k upres'd to 4k test

[Ryan Hamblin]

Did a bit of testing with the scarlet 3k... to see how bad it is to scale 3k up to 4k... if it were projected or viewed in 4k. I know that technically 5k would be better and 6k even better still... but I have to say I am very pleased with 4k/3k uprez for 4k delivery. All images are compressed jpg. I dont know a place that will host the tiffs... so if you have suggestions and want me to put those up let me know

The dividing line runs down the middle of the face... just touching the edge of the nostril on the right side. The 3k has just a hair of sharpening added.

 

[Brett Harrison]

EDIT: 3.3K anamorphic is actually slightly more pixels than spherical 4K!



A very good test Ryan, thanks.

Interestingly, Elysium was shot at 3.3K upped to 4K, and I watched it at IMAX which was pretty eye-popping.

What cameras were used to shoot the film?

Some parts of the film were shot with the Canon 5D and helmet-cams, but most of it was shot anamorphic on the Red Epic. The Epic is a 5K camera, but Elysium shot 3.3K in the end. The scope aspect ratio is 3296×1350 [squeezed down from 3296x2700 when recorded anamorphically] so we worked in 3.3K and upres'd it to 4K. That was a pretty small step up, and it shows up very well. It was amazing on a huge screen — I'm quite impressed with Red at this point, and what they're able to do with their cameras.

http://www.studiodaily.com/2013/07/how-elysiums-4k-di-integrated-editorial-and-vfx/

 

[Will Keir]

Why would they shoot Elysium in 3.3k?

A very good test Ryan, thanks.

Interestingly, Elysium was shot at 3.3K upped to 4K, and I watched it at IMAX which was pretty eye-popping.

What cameras were used to shoot the film?

Some parts of the film were shot with the Canon 5D and helmet-cams, but most of it was shot anamorphic on the Red Epic. The Epic is a 5K camera, but Elysium shot 3.3K in the end. The scope aspect ratio is 3296×1350 [squeezed down from 3296x2700 when recorded anamorphically] so we worked in 3.3K and upres'd it to 4K. That was a pretty small step up, and it shows up very well. It was amazing on a huge screen — I'm quite impressed with Red at this point, and what they're able to do with their cameras.

http://www.studiodaily.com/2013/07/how-elysiums-4k-di-integrated-editorial-and-vfx/

 

[Noah Yuan-Vogel]

Why would they shoot Elysium in 3.3k?

It was anamorphic, and 3.3k Anamorphic (8.9MP) actually uses more pixels than 4k spherical (8.8MP). So that is not a good comparison to 3k spherical which uses about half as many pixels as either of those formats...

 

[Ryan Hamblin]

Correct 3.3k anamorphic is a bit more than 3k spherical... Not drastic though... This was mainly a test to see how I would feel about using 4k if I had to use for projection.... I walked away very happy. I think if I had to print 3k it could hold up for magazine for a single page. And of course for billboard... 3k is just at 5 mega pixels

 

[William Dempsey]

Hi Ryan,

Can you clarify for me exactly what you did in this test. The lineup of the 2 halves of the frame appears to be too perfect to have been shot in 2 different passes at differing resolutions.

This leads me to think that it was shot at 4k and half the frame reduced to 3K and then uprezzed back to 4K. The potential issue with this methodology, if I am correct in my analysis ,is that the test would not accurately represent shooting originally at 3K since the effect of the OLPF would not be accurately account for. My experiments shooting on a Scarlet suggest that the OLPF is better optimized for higher res shooting. As a result a 4k original, reduced to 3k ,will give better results than a 3K original.

When I got my camera originally I was shooting at both 3k and 4k initially. After comparing the results the quality difference was substantial enough that I decided to shoot 4K on most things even when deliverables were lower res. Now I would still choose to shoot 3K in certain circumstances , but if I need high frequency detail I'll be shooting 4K.

Anyway please let us know what procedure you used to shoot this test.

Thanks,

Bill

 

[Ryan Hamblin]

No it was infact 2 different passes... I just shot them within 30 seconds of each other and had the subject hold very still. The only additional thing done to the 3k was had some sharpening. The 3k is blown up a bit more than the 4k in all the crops because hit had to be scaled up initiall to match the frame size of the 4k and then I cropped in on that. So from my testing I am seeing that 3k is very very solid for blow up to 4k and delivery. I used to be sceptical for when I was doing slomo to cut into my 4k... I no longer have a single concern.

Hope this helps understand the test a bit more

 

[Mike P.]

Two things:
1.) The 3k footage (with a touch of sharpening) actually looks like it has more texture -- "organic looking grain", as it were. Subtle filmic thing.
2.) At 3k you can shoot at higher compression ratios which could compensate for the resolution loss (and/or help in low-light, and/or reduce the above mentioned "organic grain" induced by the sharpening.)

If anything, all this test shows me is that resolution isn't everything. A lower amount of bigger, more sensitive photosites can likely be upscaled even better (presuming their cleaner).