Compression Test Results

[Zeb B]

Please point me to the thread where someone has done comprehensive Epic tests at all the RED Compression levels. I am interested in seeing at what point compression artifacts start to be and issue since shooting at 5K high frame rate 12:1 is the only option.

Thank you!

 

[Dan Kanes]

I shot an entire Fashion Identity campaign at 12:1 high speed. Will post it up here as soon as it has the go-ahead from the clients. 12:1 is not the only option for high speed, you can shoot at other ratios. I chose 12:1 because I knew we wouldn't have a lot of time for reloads where we were shooting. The results speak for themselves (I think)

 

[John Marchant]

I've tested compression comprehensively, to conclude that largely the image with which you present the sensor is almost as great a determining factor below 6:1 as anything else. 5:1 and better works almost flawlessly even with very tough images. And in my testing, what turns out to be tough on the codec isn't always easy to predict. One thing is for certain - resolution hides compression very effectively when downsampling for delivery. I'd take my chances on 4 or 5k highly compressed images far more readily than with 2k where I want to be at 5:1 or better for anything complex.

No substitute for your own tests, of course, but do bear in mind there are positions on the dial other than spinal tap '11'. Overcranking but to a slightly lesser degree gives you far better compression options if your images are testing. Check out the spreadsheet I made a few firmware builds ago (I believe its still current) http://www.reduser.net/forum/showthread.php?62183-Redcode-Framerate-maxima-spreadsheet

 

[Zeb B]

Thanks guys, my primary interest is going to be tests done at 8:1 through 12:1 since I know anything 7:1 and better is going to be pristine. On Epic using 5K at high frame rates you're forced into 8-12:1 compression so that is where the quality questions is for me. Would love a Zucato Shoot Out style test series (technical charts and subjective criteria) on the compression levels of Epic.

 

[conrad gaunt]

with the exception of certain natural scenes .. especially "tree bark" heavy scenes.. i always found rc36/ 42 overkill in regular, indoor/urban locations.

often you're just capturing extra "noise floor".. converting pixels to wavelet coefficients then compressing/"throwing away small values" is a fairly good noise reduction system and results in a more filmic/ slightly softer look.

i still love rc28, and it moves through post a little quicker

more isn't always better..

 

[Marcos Montenegro]

Thanks guys, my primary interest is going to be tests done at 8:1 through 12:1 since I know anything 7:1 and better is going to be pristine. On Epic using 5K at high frame rates you're forced into 8-12:1 compression so that is where the quality questions is for me. Would love a Zucato Shoot Out style test series (technical charts and subjective criteria) on the compression levels of Epic.

Josh:

you mean Zacuto, right?

 

[Zeb B]

with the exception of certain natural scenes .. especially "tree bark" heavy scenes.. i always found rc36/ 42 overkill in regular, indoor/urban locations.

often you're just capturing extra "noise floor".. converting pixels to wavelet coefficients then compressing/"throwing away small values" is a fairly good noise reduction system and results in a more filmic/ slightly softer look.

i still love rc28, and it moves through post a little quicker

more isn't always better..

New to RED: is RC28 a compression ratio on Epic? I see just "X":1 style choices

 

[Dan Kanes]

RC28 is on the Red One. That was the lowest setting offered on the One.

 

[Zeb B]

So nothing on higher compression testing on Epic? Seems strange

 

[Kemalettin Sert]

i tested 18:1 on 5K and it looks like shi... never went above 10:1 after that