Quantel iQ and RED .R3D

[Todd Jaspers]

Looks like Quantel will have a pretty awesome quality when it comes to RED importing..

http://www.quantel.com/page.php?n=Working_with_RED_files

[Gunleik Groven]

That looks nice...

[Patrick Tresch]

Thanks for the link.

Patrick

[Sanjin Jukic]

Looks like Quantel will have a pretty awesome quality when it comes to RED importing..

http://www.quantel.com/page.php?n=Working_with_RED_files

We shouldn't forget that Quantel is one of the platforms that is chosen by Peter Jackson's main post studio in NZ.

That's a logical move.

[Deanan]

Quantel is comparing a 4k debayer scaled to 2k with the faster/smoother direct 2k debayer. As we've outlined before for 2k output:

Best quality: 4k scaled to 2k
Better/faster: 2k high/smooth debayer
Good/fastest: 2k normal (perceptually sharper than 2k high but more aliasy)

The sharpness you get out of the 4k to 2k is from the scalar they're using.
Sometimes you want a sharper one and sometimes you want a smoother one which is why there's usually a choice of resampling algorithms.

[Graeme Nattress]

Also, if you're using the 2k high from 4k in RedAlert etc., then adjusting the OLPF compensation adjusts the downsample filter from soft through to sharp - it's up to you how you want it to look.

Graeme

[Simon Blackledge]

So they are comparing a 4k extract scaled to 2k (theirs ) to a red extract at 2kHigh (redAlert) ?

:-/

s

[Graeme Nattress]

The underlying code across RA, RC, SDK is all the same. So they could be using any of these tools. To me, and I don't know for sure, it looks like the 2k high with OLPF comp off. Putting it up high on 2k High adjusts the downsampling filter to make a much sharper downsample and does look rather nice.

Graeme

[Simon Blackledge]

Only just had my coffee so excuse the stupidity :-/

So the "soft" image was you think 2k high with olpf off

And it was just your suggestion that it should (oplf) be set high.. not what they have done.

So not a fair comparison really..

[Graeme Nattress]

That's my guess. Maybe they didn't know it effected the downsample filter.

Graeme

[Simon Blackledge]

:) cool.. understand..

Thanks Graeme

s

[damonha]

Hello all,

Damon here from Quantel.

As with many manufacturers we are still learning about Red and the different workflows that people are using with it. We want to be able to provide the best solution for our customers.

The comparison shown on our website was between importing 4K to 2K using the Red SDK decode mode and the same using our own hardware resizer (i.e. Red decode mode left at 4K with our resizer set to half of source clip). Both were imported with OLPF turned off. I have tried the same with OLPF turned to high and both pictures get better (i.e. sharpened), but the relative difference between the Red decode mode and Quantel resized picture remains the same.

One of the reasons we have taken this approach in providing the choice to use our own resizer is to provide an HD or indeed any other format workflow using Red material. For example 4K or 3K can be shot and imported (and soon conformed) into the system as true 1920x1080 either as letterboxed or as full frame (cut) mode. Importing material using the Red decode size at 4K and using our hardware sizer to make the 2K may be a little slower, but the quality is far better. We are simply providing choice to the end user which has to be a good thing.

We are certainly not trying trying to pull the wool over anyone’s eyes just providing more options to our customers when working with Red material.

Quantel are very much aware of what impact Red is having in the marketplace and are working hard to provide the best workflow solutions we can.

Many thanks,

Damon Hawkins
Quantel.

[Graeme Nattress]

Damon, this is getting confusing. The "half" decode mode goes direct from the raw data to an RGB decode of half of the native pixel dimensions. The downsampling filter used needed a control for effectively what type of anti-alias filter is used. For reasons of keeping things simple it got tied to the OLPF compensation.

The half mode is designed for both speed and quality, but you do need to set the OLPF comp up high to see what it looks like, or else the downsample filter will approximate to gaussian and be soft. Adding the OLPF control does not post sharpen the 2k at all, but just adjusts the coefficients of the downsample filter, and hence doesn't take any longer to render either.

Better results can always be obtained from a full adaptive 4k demosaic followed by the downsample of your choice, in whatever tools you use to downsample in.

So, if you're going to publish a comparison, it would look more appropriate if the RED half decode was shown at either a variety of OLPF settings, or at least just show the most comparable one for your example.

Thanks,

Graeme

[damonha]

Hi Graeme,


Better results can always be obtained from a full adaptive 4k demosaic followed by the downsample of your choice, in whatever tools you use to downsample in.

I understand and agree.

We are simply offering a choice. If the user wants to work at 2K from material shot at 4K then they can use our resizer (which is a realtime process) post the 4K demosaic. If not they can use the Red downsample, which whilst faster, obviously won't be of the same quality. The OLPF can be used off or on high in either case without any impairment to speed.

One of the main reasons for introducing our own hardware resizer on import and conform is that it also gives choice and opportunity for an HD, or indeed any other resolution workflow with Red material.

We are just trying to provide flexibilty.

Cheers,
Damon.

[Graeme Nattress]

As are we, Damon. Flexibility is good. However, the order of magnitude of the speed improvement of the half rez demosaic over the full rez is significant, but the quality difference of a full rez demosaic and downsample compared to direct half rez demosaic is much less than your image example implies.

The web page says "first using the RED sizing", but there is not one "RED sizing" due to the choice of downsampling filter being a user choice set through the OLPF control. When people see that, they must be thinking "my is that RED sizing soft", when although that is true, it's only because it's on a "soft setting" rather than a "sharp setting".

If you can send me the R3D file (great test shot BTW), I can run through the half demosaic sharpness settings and then we can both educate our users about the flexibilities both the RED tools and Quantel offer.

Thanks,

Graeme

[Patrick Tresch]

Is this the way your both companies work together?

Through a formu?

????

I don't have anything against new ways of communications but as an newbie and end end user It just makes wonder.

[Simon Blackledge]

But... Always a but sorry.

Are we not therefore comparing a full 4k extract that is post scaled to a red extract at 2k?

Yes the user gets more flexibility post 4k. But then you have spent longer extracting the 4k rather than 2 no?

Ie a 4k extract scaled to 2k in after effects would be sharper.

Have I got this wrong then please correct me.. Were all learning.

[markh]

The team here at Quantel have just annouced the beta phase of Red conform for eQ, iQ and Pablo. Details are in the white paper at http://www.quantel.com/list.php?a=Library&as=White_Papers

As the thread suggests we're learning here too so all comments or questrions very welcome.

Mark

[Fredrik Callinggard]

Is this the way your both companies work together?

Through a formu?

????

I don't have anything against new ways of communications but as an newbie and end end user It just makes wonder.

I think it's great, let's us learn in the process.

[Scott Simmons]

Is this the way your both companies work together?

Through a formu?

????

I don't have anything against new ways of communications but as an newbie and end end user It just makes wonder.

that's a good statement. Hopefully RED and Quantel have a closer working relationship than this discussion board.

[markh]

Anyone in the Phillipines interested in what Quantel are doing is welcome to come see us in Manila in the first week of January. Details here

http://www.quantel.com/list.php?a=News&as=Events