Chromaticity diagram

[leoenc]

Hi

Does anyone know of a tool, preferably a filter for avisynth or ffmpeg, that can analyze and build a CIE 1931 chromaticity diagram from a video file?

I would like to verify that a video indeed uses the color space stated in its metadata.

[Selur]

Searching in the ffmpeg filter documentation -> https://ffmpeg.org/ffmpeg-filters.html#ciescope

[richardpl]

I'm really puzzled, how you are supposed to know what real color space is in video, when video does not use XYZ pixel format from beginning.

[Selur]

https://github.com/Lypheo/Color-Tools for VapourSynth might also work. (sadly no binary)

[leoenc]

@Selur Thank you very much, I will look into both these options.
@richardpl I don't. I was thinking that a chromaticity diagram would allow me to visually estimate what color range the video is using. Hopefully, if I was to use the ffmpeg ciescope filter on an HDR file, I would set the color system to rec2020 and roughly compare the resulting triangle with a correct rec2020 diagram.

[wswartzendruber]

Hi

I would like to verify that a video indeed uses the color space stated in its metadata.
It sounds like you want to use that color space's primary coefficients, convert to XYZ, and then look at the output to see if it makes sense.

I would think that creative intent would throw that sort of assessment off completely. While this might work for The Avengers, it likely won't work for Zack Snyder's Justice League.

[leoenc]

[QUOTE=wswartzendruber;1963999]It sounds like you want to use that color space's primary coefficients, convert to XYZ, and then look at the output to see if it makes sense.

So, if I understand correctly, I could feed an HDR video with AVS ConvertYUVtoXYZ into ffmpeg's ciescope filter.

[richardpl]

You could see real gamut of video only and only if input to filter is in XYZ colorspace.

[leoenc]

You could see real gamut of video only and only if input to filter is in XYZ colorspace.

So if I'm using ConvertYUVtoXYZ, input to ciescope filter should be in XYZ colorspace.

[wswartzendruber]

So if I'm using ConvertYUVtoXYZ, input to ciescope filter should be in XYZ colorspace.
In theory, this should work. Make sure the ConvertYUVtoXYZ function is aware of the primary coefficients of the source colorspace. Obviously, ciescope must expect XYZ input.

[wswartzendruber]

Remember, though, that creative intent can throw this sort of thing off:

https://th.bing.com/th/id/R.1cbedb85c00c5297ad72a34895f1aa5b?rik=yTQTjfEPtEUKZA&riu=http%3a%2f%2ftheactionelite.com%2fwp-content%2fuploads%2f2021%2f03%2fzack-snyder-s-justice-league-review-1200.jpg&ehk=NhBUPkFPwVevqg7%2fDvFST0bGkLOzYSfWrmF3zmC5vCQ%3d&risl=&pid=ImgRaw&r=0