Blackmagic Forum

Tom wrote:

TZuck wrote:ProRes 4444? this would be nice if the sensor was capable of doing 4:4:4 chroma (no)subsampling.
here's an article that explains what I mean:
http://www.xdcam-user.com/2011/02/when-is-444-not-really-444/

so to recap, if the BMCC were to record ProRes 4444 (really it should be ProRes 444, unless the camera can film transparency) it would be 99.99% same as ProRes 422 HQ, but eat up more space on drives. But if you really wanted to eat up more drive space then film in RAW, because at least that give you more bit depth to play with in post. If the point is to improve the BMCC I don't see how adding ProRes 4444 does that, same with Cineform RAW, I don't understand why you would film in it.

1: the sensor does not perform any chroma subsampling.
2: Bayer pattern sensors and chroma subsampling are not the same thing and should not be directly compared. Similar ideas, but very different thing with different implications.
3: Cineform RAW would be beneficial because the data is bayer data with very good, visually lossless compression with a high colour bit depth. It means the benefits of not having pre-debayered video footage such as with DNG, the benefits of high colour bit depth such as with DNG, but with much smaller files - and of course all the cool things you can do with cineform such as with their meta data.

1.) It would appear as though i've been mistaken. gone and messed up my words...
2.) correct. that being said ProRes 444 still should not be noticeably better quality coming from the current camera. and if you want a true 4:4:4 image from this camera the resolution would be 1200 x 675. Canon has this on the c300 for 1080p only because the sensor is 3840 x 2160. The point of a bayered sensor is to debayer it. software is still guessing at what was there, and any other way to debayer is simply preferential, which is your point in #3? that you can decide how it gets debayered?
3.) You've made valid points. carry on while i go and educate myself more.

TZuck wrote:

Tom wrote:

TZuck wrote:ProRes 4444? this would be nice if the sensor was capable of doing 4:4:4 chroma (no)subsampling.
here's an article that explains what I mean:
http://www.xdcam-user.com/2011/02/when-is-444-not-really-444/

so to recap, if the BMCC were to record ProRes 4444 (really it should be ProRes 444, unless the camera can film transparency) it would be 99.99% same as ProRes 422 HQ, but eat up more space on drives. But if you really wanted to eat up more drive space then film in RAW, because at least that give you more bit depth to play with in post. If the point is to improve the BMCC I don't see how adding ProRes 4444 does that, same with Cineform RAW, I don't understand why you would film in it.

1: the sensor does not perform any chroma subsampling.
2: Bayer pattern sensors and chroma subsampling are not the same thing and should not be directly compared. Similar ideas, but very different thing with different implications.
3: Cineform RAW would be beneficial because the data is bayer data with very good, visually lossless compression with a high colour bit depth. It means the benefits of not having pre-debayered video footage such as with DNG, the benefits of high colour bit depth such as with DNG, but with much smaller files - and of course all the cool things you can do with cineform such as with their meta data.

1.) It would appear as though i've been mistaken. gone and messed up my words...
2.) correct. that being said ProRes 444 still should not be noticeably better quality coming from the current camera. and if you want a true 4:4:4 image from this camera the resolution would be 1200 x 675. Canon has this on the c300 for 1080p only because the sensor is 3840 x 2160. The point of a bayered sensor is to debayer it. software is still guessing at what was there, and any other way to debayer is simply preferential, which is your point in #3? that you can decide how it gets debayered?
3.) You've made valid points. carry on while i go and educate myself more.

Pretty much bang on,

basically, the effect that applying chroma sub-sampling to video has is fairly uniform, whereas depending on how you debayer the sensor data - the variation and quality of image can vary in massive ways! (http://blog.elphel.com/2010/11/zoom-in-now-enhance/ scroll down to "PRELIMINARY RESULTS")

I agree that Prores 4444 is fairly pointless for the BMCC. I would rather 12 bit compressed video at 4:2:2 than 10bit 4:4:4 from this camera. But really, Cineform RAW would be perfect for this camera, it would be like the best of both worlds.



On a slightly different note,
When John Brawley released both ProRes film and RAW clips from pool shark - I did a quick grade, applied the exact same settings and tried to match the clips perfectly, exported them both as Cineform444 in 1080p - did a blind test with 3 friends (not cinematographers or film people, just your typical joes) and told them that one was from RAW 2.5k and one was shot in HD.

They all thought that the ProRes looked sharper and was probably the Raw one. Go figure! haha

I have also tried encoding 4:2:2 and 4:4:4 clips from Alexa footage, even when pixel peeping, no one could tell the difference. when enlarged to about 400%, we could just notice a slight difference on a strand of hair!


I am not saying that 4:4:4 is pointless, just that there are probably far more important considerations when it comes to the final quality of the video. (not saying anyone has said otherwise)

Whilst there might not be a visible or discernible difference between 444 and 422 there is a big difference when you're trying to grade it. ( assume there's 444 worth of info ere to start with )

JB.

John Brawley wrote:Whilst there might not be a visible or discernible difference between 444 and 422 there is a big difference when you're trying to grade it. ( assume there's 444 worth of info ere to start with )

JB.

Very true! thank you for raising that point!

I would love it if the camera could write ProRes 422 at the full resolution of 2400x1350. The extra resolution would still be there for re-framing or stabilisation, but you would have a smaller file size.

+1