ср, 4 сент. 2024 г., 20:31 Andrew Randrianasulu <[email protected]>:
ср, 4 сент. 2024 г., 20:26 Phyllis Smith <[email protected]>:
On Tue, Sep 3, 2024 at 3:41 PM Andrew Randrianasulu via Cin < [email protected]> wrote:
found via ffmpeg bugtracker referencing our bugtracker ;)
https://trac.ffmpeg.org/ticket/6793
Interesting. So the problem is with the Sony XAVC format (used by at least FDR-AX100)? And since the ffmpeg bug report is still open, but it is old, does that mean it is not fixed? And it is not a CinGG bug, but an ffmpeg bug?
As far as I understand this bug about specific timecode in this file. So, it reads ok (?) just not with all info.... I tried to use it for testing, assuming xvycc really should be compatible with yuv /bt709 signal paths assuming there was no clipper.
It seems to.work in sense image looks ok, but it looks a bit differently if I add F_colorspace filter with xvycc transfer function.
10 days happen I found article describing decoding a bit more https://www.researchgate.net/publication/250999893_192_xvYCC_A_New_Standard_... and there was half of iec-61966-2-4 https://cdn.standards.iteh.ai/samples/14431/9e5a80b3c1a54ad0866f818def33466c... describing xvycc conversion to cie xyz 1931 via RGB It seems that decoder need to apply some 3-segment curve to get RGB values outside of [0,1] and then those can be manipulated as usual? "NOTE The possible range for non-linear R’G’B’ (709) calculated from, for example, equation (11) will be between -1,1206 and 2,1305" so, it seems to use really aggressive optico electrical transfer function coefficients around zero , 4.5!