Question about cineon to linear

[esi]

Hi all,

I know this is not about VRAY, but there are a few people here who use Fusion and combustion and NUKE, and I have some question regarding this colour convertion.

Basically, here it is.

I have new compositor in the studio and we have been talking about, in his case combustion, and overall what technically happens to the Cineon LOG file.

The issue is hard to talk about, because the software he is using is not working with the information stored in this log space correctly, on the other hand in Fusion and NUKe it is reading and converting correctly. So I am wondering. Can some one explain the technical convertion that is happening in the process.

The process in question is taking a cineon log file converting it to a EXR file 32 bit or 16bit float, does not matter to me which.

This is what I know. The negative film is scanned and placed into the cineon log file, the entire exposure that was intended is captured. If you had your choice you would want to colour correct in Log due to this range for example. However we all know that all of our work is linear, many compositing processes expect linear values etc. There is no stead fast rule mind you, but it seems be a way of working as computers get faster etc.

We then need to convert the intended exposure range black and white points (standard is 95-685) as established by Kodak. You can change this so you can get every little value above this or below, but I am just working under the standard way for this discussion.

We then remove the log fuction from the cineon file and transfer it to linear exr float. This will have enough headroom to maintain what was captured in the log file, but allow us to work in a linear way which makes me happy for many reasons.

Now, one question that has come up, is CC or linear or not to CC in linear. My personal opinion is this: Log would be best, but since in float all the info is there it should not matter. Or if you can't work this way, linear, Log would be best.

Second question, to Key in Log or to Key in linear. My opinion on this is to Key is Linear due to my comment on compositing software. the log image in its native format is not telling the compositing software what the true value is there. correct? or false?

Finally, general compositing with log or linear. Basically since all transformations etc are done in a linear way working in log seems to make no sense. True or false?

Also, for those combustion people out there, when you use the bit depth converter, it does not seem to convert from log to linear correctly. It seems to clamp the values to 1. It does not preserve the dynamic range. Can anyone confirm this?

Well thanks for anyone who can help.

[panthon]

Re: Question about cineon to linear

Hi all,


Also, for those combustion people out there, when you use the bit depth converter, it does not seem to convert from log to linear correctly. It seems to clamp the values to 1. It does not preserve the dynamic range. Can anyone confirm this?

Well thanks for anyone who can help.

I am no expert at all but it seems you dont need to transfer from cineon to exr in combustion. As i can read:

When you import footage, the bit depth is initially set to the bit depth of the footage in the Output controls of the Footage operator. For Cineon files, you can make combustion automatically convert the files from logarithmic data to linear data and set the linear bit depth. You can leave the bit depth at 10 bits, or change it to 8 bit, 10 bit, 16 bit, or Float.

To automatically convert Cineon files to linear:

Choose File | Preferences or press Ctrl+; on Windows, or choose combustion | Preferences or press command+; on Macintosh.
In the Footage category, click General.
The Footage preferences appear.

Enable Log to Linear Conversion.


Select a bit depth from the Conversion Depth list.
Click OK.
Now, if you import a Cineon image sequence, and look in the Output controls of the Footage operator, Use LUT is enabled, and the Depth list is set to the Conversion Depth in the preferences. If you show the LUT Editor controls, you see the conversion method is set to Log to Linear.

Since a LUT can be saved inside a Cineon file, the Log to Linear method uses the LUT that is in the file, if available.

I hope it helps. I am using now a 10 bit format too, but it comes from dvcpro 50 and it's showing correctly in combustion. I seriously doubt that dvpro50 4:2:2 can have such dynamic range as a cin file from film...

Alternatively, u can skip the setting in the preferences and check the option "use lut" under footage controls, output. There you should use the option log to linear and adjust the values. AFAIC, this is only for viewing purposes, so depending on what format you are intending to output to, you could end up having to readjust the resulting render later on.
If you use some CC operators in combustion, this information will be "baked" into the footage. In either case i have not seen any clamping of the data, but i have no access to cin files so i can't tell for sure.

[esi]

Thanks for the reply Panthon,

Yeah you need to use a cineon file to understand fully what I am talking about.

Technically the clamping that I am talking about is when you bring in a cineon file with values higher than one. You can try this with the marcie cineon file that has plenty of extra data to play with off the kodak site. You tell it to use a particular exposure range and then where the white point is you are then clamping it. As it is making any value above that white point a value of 1 in float. I just think it should be more like Nuke and keep all the dynamic range in tacked and allow the user to decide when and how.

Just remember your ten bit and the cineon 10 bit are the same in this regard, but after that they are different. Mostly because there is a log function applied to it to allow for the dynamic range of negative film as scanned into the computer.

The issue is really about how to get the log removed and then working in a linear way without any fear of lossing data. Now I tryed to explain above that I know the process but our new compositor is new to this process and is having a hard time understanding the process. He is under the understanding that this process of capturing the exposure range in the convertion tool is indeed making it a 8 bit image. I have tried everything I know to convince him and now I wanted to post these questions to get a more simple way to explain.

In my previous post these are some of the questions asked by the compositor and what my answer are.

If anyone here like Chris is reading this, a thread about this would be very helpfull for this person. Not to mention anyone else wanting to understand.

thanks in advance,

[emora]

Hello Mark..
I've used combustion a lot int he last months to composite 2k cineons without a problem, there are many aspects as you point out, but the basic workflow that has worked for me is: dont use the bit converter, there is basically no need unles you want keep your footage at another bit depth..or ar mixing it with other sources at different bit depths..
the proper way to treat film scans footage with combustion ( Cineons, DPX, or EXR's, etc) is load the footage normally, and use a LUT to do all your composite, what the LUT is doing is "selecting" only a range of color from the full range stored in the cineon so you can see it in your monitor.. combustion has good "average" default straight LUT, but you can use Log to linear or linear to log, or import any other LUT say.. 12log to 10, etc...
The LUT can even come from a color space/range that Nuke exports..and then imprt into Combustion..but I'm almost sure the ones inside combustion sholud work.. about your question.

Second question, to Key in Log or to Key in linear. My opinion on this is to Key is Linear due to my comment on compositing software. the log image in its native format is not telling the compositing software what the true value is there. correct? or false?

I'll say False..
The software does "see" all the color depth inside the cineon.. is your monitor that cannot display them.. thats way a LUT is necessary to work inside a range that you can see.. to test this.. load a cineon secuence and with the color corrector, sample any high value, youll see that goes way beyond 255.. all that extra info is there..
I'm guessing you are compositing CG elements over the cineon's..
so render to EXR or Cineon.. with all the extra color data that the format permmits.. then comp inside Combustion using a LUT and all should be good..
now what LUT to use?.. that is another subject.. it can come from another software, so all composites are done in the same color range.. it can come from the people doing the DI... or you can create yours based on the shots you are doing.. as long as you end up with a renderd file that has the extra info to tweak..

hope it helps

Esteban Mora
Citrica Studio

[esi]

Thanks for the notes emora,

Bascially I agree with you in regards to the viewLUT or LUTs in general. However we are moving towards a linear float workflow. You can say that if you load a Cineon into a Float environment and do your log to lin you will have the same information that you will have in a cineon log file. However when you work in a 10 bit or 16 bit everything above the white point gets clipped and when you do colour work and transformations etc you will loose that detail. In most cases this is fine, becuase it falls in the exposure range, however we go to a DI process so when we are finished working in float we render out again in cineon log etc and everything we have done is still in tacked.

As far as the keying I am still out on this, keying software to my knowledge wants to see linear information. You can do it in log and see the LUT results, but I think this is just making it harder since the software wants to deal with linear. Even in the shake manual they go into great detail on this.

Anyway thanks for thoughts.