Blackmagic Forum

Is there an official specification of the Blackmagic cameras' colorimetry for ProRes output when Film mode is selected? Specifically, I mean the RGB primary chromaticities, the RGB transfer characteristics, the R'G'B' to Y'PbPr transformation coefficients, and the output ranges for the 10-bit Y', Cb, and Cr channel values prior to ProRes encoding.

This information is required to properly convert the decoded Y'CbCr values back to RGB and interpret them.

Also, I've noticed that the Cb and Cr channels seem to use less than half of the available 0-1023 channel value ranges, resulting in pictures with low saturation and effectively wasting 1 bit of precision in each sample. Why is this? I can think of no good reason for it. The Y' channel appears to be using the normal rec.709 Y' channel range of 64 to 940 (or close to it).

The low contrast appearance of Film mode footage is a completely separate issue from the low saturation: it's a consequence of taking the camera's log-transfer large dynamic range and displaying it in some smaller dynamic range display-referenced standard like rec.709 or sRGB without properly mapping the levels.

Few are technically knowledgable enough to give a coherent reply.
I was unaware prores film mode was not rbg...

paulkosmala wrote:Few are technically knowledgable enough to give a coherent reply.
I was unaware prores film mode was not rbg...

The full name of ProRes (HQ) is ProRes 422 (HQ) which implies Y'CbCr coding.

balazer wrote:The low contrast appearance of Film mode footage is a completely separate issue from the low saturation: it's a consequence of taking the camera's log-transfer large dynamic range and displaying it in some smaller dynamic range display-referenced standard like rec.709 or sRGB without properly mapping the levels.

Log output from most other cameras (Alexa, F65, C300 etc.) is also wide gamut, and therefore appears desaturated prior to colour matrixing. This means it should not be expected to use the full range of the Cb and Cr channels. This could be regarded as "wasting" bits, but it could also be considered as a contributing factor to compression efficiency.

balazer wrote:Is there an official specification of the Blackmagic cameras' colorimetry for ProRes output when Film mode is selected? Specifically, I mean the RGB primary chromaticities…

To be pedantic about the original question, a camera does not have RGB primary chromaticities. It has overlapping spectral responses in the RGB photosites. The camera's colour matrix converts this into something approximating RGB primaries.

I believe that the details of this process are not publicly published at this time.

Nick Shaw wrote:Log output from most other cameras (Alexa, F65, C300 etc.) is also wide gamut, and therefore appears desaturated prior to colour matrixing. This means it should not be expected to use the full range of the Cb and Cr channels. This could be regarded as "wasting" bits, but it could also be considered as a contributing factor to compression efficiency.

Are the Blackmagic cameras wide gamut, and what is the gamut? I've never heard anyone say they are wide gamut.

The desaturated appearance of F65 wide gamut footage when viewed without proper gamut conversion is in no way a waste of the ranges in the Cb and Cr channels. The full ranges can be used by colors in the extended parts of the gamut. I see no evidence that that is the case for the Blackmagic cameras.

To be pedantic about the original question, a camera does not have RGB primary chromaticities. It has overlapping spectral responses in the RGB photosites. The camera's colour matrix converts this into something approximating RGB primaries.

I believe that the details of this process are not publicly published at this time.

No matter how you get from the RGB photosite sample values to the RGB primary values, you need to know the chromaticities of the RGB primaries if their values are to mean anything. Rec.709 and every other video standard specify the primary chromaticities. My guess would have been that the Blackmagic cameras use rec.709 chromaticities also.

balazer wrote:The desaturated appearance of F65 wide gamut footage when viewed without proper gamut conversion is in no way a waste of the ranges in the Cb and Cr channels. The full ranges can be used by colors in the extended parts of the gamut. I see no evidence that that is the case for the Blackmagic cameras.

The F65 certainly has a wider gamut than the BMCC, but I have not seen F65 footage which uses the full extents of the Cb and Cr channels. But I have not done tests to specifically look for that.

balazer wrote:No matter how you get from the RGB photosite sample values to the RGB primary values, you need to know the chromaticities of the RGB primaries if their values are to mean anything. Rec.709 and every other video standard specify the primary chromaticities. My guess would have been that the Blackmagic cameras use rec.709 chromaticities also.

Blackmagic film mode does not use Rec.709 primaries. It uses its own Blackmagic Film gamut. A matrix (not currently publicly published) will convert this to primaries which "look good" on a Rec.709 display. This is what the video mode is. This is what all digital cinema cameras do in my experience. They each have their own look, rather than being scientific measuring instruments actually measuring the chromaticities in a scene.

Absolutely. We tested Alexa, F65 and Epic against each other carefully and each has a specific look, in particular when it comes to colors. And none of these manufacturers is publishing their tricks in color matrizes.

Nick Shaw wrote:Blackmagic film mode does not use Rec.709 primaries. It uses its own Blackmagic Film gamut. A matrix (not currently publicly published) will convert this to primaries which "look good" on a Rec.709 display.

Well that's what I'm getting at. How am I supposed to convert the Blackmagic film gamut to rec.709 and get it to look good, except to use Resolve? The point of ProRes is that it's a standard that you should be able to read in lots of different applications.

Seems to me this mystery matrix should be published.

balazer wrote: Seems to me this mystery matrix should be published.

yes agreed. I've been trying to find a way to generate scene-referred linear files from the DNG sequences. We need BMD to let us know how to interpret the "film" raw files, other than to say "just use Resolve"

Gary Jaeger wrote:

balazer wrote: Seems to me this mystery matrix should be published.

yes agreed. I've been trying to find a way to generate scene-referred linear files from the DNG sequences. We need BMD to let us know how to interpret the "film" raw files, other than to say "just use Resolve"

Set your project type to ACES, your IDT to CinemaDNG and export EXR. Boom, scene linear image sequences.

The matrix required to get from camera primaries to ACES XYZ are embedded in the DNG tags.

Jacob Balazer wrote:

Nick Shaw wrote:Blackmagic film mode does not use Rec.709 primaries. It uses its own Blackmagic Film gamut. A matrix (not currently publicly published) will convert this to primaries which "look good" on a Rec.709 display.

Well that's what I'm getting at. How am I supposed to convert the Blackmagic film gamut to rec.709 and get it to look good, except to use Resolve? The point of ProRes is that it's a standard that you should be able to read in lots of different applications.

Seems to me this mystery matrix should be published.

Actually, ProRes is NOT a documented standard. It has been reverse engineered by FFMpeg and been implemented under NDA by many companies.

DNxHD and CinemaDNG are documented, open standards.

I didn't say the ProRes file format is an open standard. ProRes is common in the industry, and any app can use it through the QuickTime API.

Please stay on topic.

Jacob,

It's usage is common and wide spread, but it is not 'standard' to pull a direct quote.

On topic, the Film to Rec709 LUT is provided as a cube and shaper LUT with Resolve. That plus OCIObakelut should be enough to massage it into whatever format you need for your workflow.

I'm not interested in using Resolve to convert footage. I want to know how to interpret Blackmagic footage correctly in my own software. Can Resolve's LUTs be exported?

Balazer,

You are not reading my posts carefully enough.

Resolve ships with the LUT as separate cube files, they do not need to be exported.

If you are working with DNGs directly, the matrix required to get into ACES XYZ are embedded in the DNG tags.

I don't need you to be so condescending. I'd never heard of a cube file before.

balazer wrote:I don't need you to be so condescending. I'd never heard of a cube file before.

I don't need you to be so acerbic and demanding. Truce? Asking questions is good but tone is everything.

http://opencolorio.org/ has a set of tools for converting and manipulating lut files, that you may find useful if you application does not support cube lut files directly.

balazer wrote:I don't need you to be so condescending. I'd never heard of a cube file before.

A Cube LUT is simply a 3D LUT

Here balazer this may help inform you about LUTS without any condescending remarks, just a few simple links to read.

"Cubes may be of various sizes and bit depths. Often 17×17×17 cubes are used as 3D LUTs. The most common practice is to use RGB 10-bit/component log images as the input to the 3D LUT. Output is usually RGB values that are to be placed unchanged into a display device's buffer."
From http://en.wikipedia.org/wiki/3D_LUT
and read here
http://www.lightillusion.com/luts.html

It's worth pointing out that the LUT which is built in to Resolve includes a tonemap to fit the dynamic range of the camera into a 0-1 display range. This means it does not contain the information necessary to convert Blackmagic Film into scene referred linear light data.

But yes, with the use of the correct transfer function (you will find S-Log is not a million miles away) you can, for example, work with BM Film footage as scene referred linear light in Nuke, and use a variation of the Resolve LUT as a viewing LUT.

This is how I work with Alexa LogC and Sony S-Log(2) media, and the same principles apply for BM Film.





I assume this is the kind of thing the OP was looking to do.

(edited to show the BM Film image with no LUT as well)

Nick Shaw wrote:But yes, with the use of the correct transfer function (you will find S-Log is not a million miles away) you can, for example, work with BM Film footage as scene referred linear light in Nuke, and use a variation of the Resolve LUT as a viewing LUT.

This is how I work with Alexa LogC and Sony S-Log(2) media, and the same principles apply for BM Film.

BM_Film_Nuke1.png

BM_Film_Nuke2.png

I assume this is the kind of thing the OP was looking to do.

(edited to show the BM Film image with no LUT as well)

Thank you for your suggestion, I have same problem with linear workflow. Wonder if there is any update after four months

As a side note FFMpeg now has a LUT filter function (referred to as lut3d). It's quite new and almost certainly doesn't use the GPU. Has anyone had any luck using one of Resolve's cube LUTs with that function ?