ProRes vs. REDCODE at AppleInsider

[Drew Mylrea]

Sorry if this has been already posted in a pre-existing thread. AppleInsider (http://www.appleinsider.com/articles/07/04/18/a_closer_look_at_apples_new_prores_422_video_forma t.html) has posted a pretty awesome run-down of the ProRes422 codec and dives in to how it specifically compares with the Red's Wavelet compression. Thoughts?

edit: this should probably be moved to the workflow section :0

[Brook Willard]

You forgot to post the link. Here (http://www.appleinsider.com/articles/07/04/18/a_closer_look_at_apples_new_prores_422_video_forma t.html) it is.

It's a brief blurb on RED and it's decently accurate.

[Stokestack]

It's a decent write-up. But it would be more appropriate to compare ProRes to DV100 (DVCProHD) than Redcode, because it's a 4:2:2 codec. Redcode is supposed to be 4:4:4.

[Jim Arthurs]

A couple notes... ProRez is full raster, square pixel, therefore a better matching down-convert for REDCODE than using DVCPRO 100 of any flavor. Also it is designed to go multiple generations better than using DVCPRO 100...

[Bruce Allen]

The summary is, ProRes is incredibly similar to DNxHD. Judging by the font used in the table & graphs, I'd say they were made by Apple, not AppleInsider.

They also get REDCODE a little bit wrong - they talk as if it were a very lossy 4:4:4 4K codec. The joy of REDCODE is that it doesn't sample RGB per pixel - pointless anyway, since with a Bayer matrix, you only record Red OR Green OR Blue per pixel. REDCODE embraces and uses this fact, storing only one color channel value per pixel in order to be more efficient and fit the sensor better.

Bruce Allen
www.boacinema.com

[Damien Molineaux]

Is ProRes 10 bits after all, I read somewhere it was only 8 bits ? I hope it's 10 bits, that would be nice.

edit : Ok, I misunderstood Mike Curtis' report and thought it was 8 bits, but it is 10 bits, good !

As to Redcode, it's RAW so you can't speak of X:X:X color space until it's been debayered. See this thread on Cinematography.com :
http://www.cinematography.com/forum2004/index.php?showtopic=22304
Yes I go there sometimes and they have useful info. I just ignore those few (and they are few) that haven't grasped what Red is about yet, or maybe they'll just stick to film, hey I love film, it can beautiful. Anyways, there's some great info there, amongst which info about bayer sensors, I know there are different types, which I'm not familiar with, but from what I gathered, Bayer sensors normally have half their pixels green, one quarter red and one quarter red, so one could say it's 4:2:2 ; BUT the debayering process, if working off more than one frame, changes that. But I don't think it can give you 4K 4:4:4 however, correct me if I'm wrong. I am confident Graeme is completely on top of this, and for those who have seen the results, Redcode seems to be stunning. I hope I haven't misunderstood what I read. Debayering is pretty new to me, I'm not pretending I'm an expert (won't stop Cinematography.comers from callng me a jumblemouth however, it's their new thing).

Cheers,
Damien

[Bruce Allen]

Yes. There is no direct comparison between Bayer and 4:x:x. Probably 4:3:3 or maybe 4:2:2 depending on situation.

I am just saying that they assume incorrectly that REDCODE is 4:4:4 and that they assume incorrectly that it must use a more lossy compression. I think it just does a more clever compression because it understands the Bayer matrix.

Bruce Allen
www.boacinema.com

[Nik Manning]

DVCPRO 100 is a wonderful codec but even panasonic has moved on to something else.
Does anyone know if ProRes supports alpha channels?
Can you use ProRes in Motion 3?
Will you be able to capture ProRes with existing AJA capture cards?
Does ProRes work with G5 apple computers?
Can you use the AJA IO HD with a PC computer for capture to ProRes?
Is AJA the only card manufacturer allowed to capture ProRes?
How does ProRes compare to Black Magic Codec and Cineform codec?
Will ProRes have After Effects support or will Apple allow Adobe to work with and render to the codec?
Can you capture SD to ProRes codec?

I am sure all of these question cannot be answered now but I wanted to get them all out before I forgot them.

Cineform is seeming a tad over priced at the moment.

[Bruce Allen]

I'm sure you won't be allowed to use the AJA IO with a PC to capture to ProRes :(

In fact, there is quite a high chance ProRes won't be cross-platform, which would suck - a lot. Cineform and DNxHD look like they're finally becoming cross-platform, thank goodness...

ProRes will almost certainly work like a standard QuickTime codec, so After Effects, Motion 3, etc should definitely work.

I would love it if ProRes supported alpha but again, probably not...

Bruce Allen
www.boacinema.com

[Damien Molineaux]

<cut...

I would love it if ProRes supported alpha but again, probably not...

Bruce Allen
www.boacinema.com

Aren't alpha channels a Quicktime feature ? Should be supported.

Cheers,
Damien

[Anders Holck]

No, the codec must support it.
The Apple DV codec doesnt support an alpha channel, while the Avid DV codec does.

[Andrew M.]

Yes. There is no direct comparison between Bayer and 4:x:x. Probably 4:3:3 or maybe 4:2:2 depending on situation.

I am just saying that they assume incorrectly that REDCODE is 4:4:4 and that they assume incorrectly that it must use a more lossy compression. I think it just does a more clever compression because it understands the Bayer matrix.

Bruce Allen
www.boacinema.com

Unverified:

RED RGB will be 4:4:4
It mean that debayer algorithm will provide unique R,G,B values for every pixel out of the 4096 pixels of the horizontal line and whatever vertical for given aspect ratio.

Now how it will compare to idealistic RGB 4:4:4 it depends from many factors:
Sensor itself
DeBayer algorithm
Oversampling factor (how many sensors each 4096 pixel line will have)

Remember that 3XCCD sensor that some will call idealistic RGB 4:4:4 are almost at par when compared with Bayer sensor.
Some factor that degrade 3XCCD sensors quality versus Bayer sensors are:
Prism aberration
3 sensor misalignment
Lack of light uniformity in 3 sensors

see Chroma subsampling:
http://en.wikipedia.org/wiki/4:4:4

Conclusion, RGB 4:4:4 from 10% oversampled 4K format from Bayer sensor and not oversampled 3X4K-CCD sensor is comparable.

[Bruce Allen]

Not all Quicktime codecs support Alpha. H.264, Sorenson, DV, for example

Bruce Allen
www.boacinema.com

[Bruce Allen]

Andrew, I think you could be right on the money about output quality and that 3-CCD "4:4:4" 4K sensors might not yield better image quality. Also, that 4:x:x is getting to be something of an outdated term with the explosion of Bayer-sensor cameras.

Here's a nice time piece - Graeme's original 2005 LAFCPUG article:
http://www.lafcpug.org/Tutorials/basic_chroma_sample.html

At the end, he has examples showing how his G Nicer plugin uses luma info to improves on 4:1:1 to output something closer to 4:2:2 in quality... nice.

Cheers

Bruce Allen
www.boacinema.com

[Craig W. Bickerstaff]

What ever happened to these 3 layer CCDs that their using in modern Digital SLR cameras.

[GlennChan]

You mean Foveon? Do a search, it's been discussed a lot. The short story is that it is too noisy / has too low exposure latitude.