4 ms·
There is no reason why YCbCr should be visually worse than RGB if the conversion is accurate
by WithinReason 2y ago
There is no reason why YCbCr should be visually worse than RGB if the conversion is accurate
- meindnoch 2y agoIt depends on the exact meaning of "YCbCr" and the meaning of "RGB". Is it BT.601? BT.709? BT.2020? Adobe RGB? Display P3? Also extra fun is guaranteed if one end of the video cable is encoding with e.g. BT.601 primaries, while the other end is decoding as e.g. BT.709, or vice versa.
- WithinReason 2y agoRGB to YCbCr conversion and back is well defined, and RGB->YCbCR->RGB should give you back the original RGB values, no matter what the colorspace was.
- meindnoch 2y agoLossless roundtrip conversion is only true for schemes like YCoCg-R [1]. It is not true for Rec.601, Rec.709, etc. because these standards require quantizing to 8bits (actually even less than 8bits due to the 16-235 thing). [1] https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/2008_ColorTransforms_MalvarSullivanSrinivasan.pdf https://www.microsoft.com/en-us/research/wp-content/uploads/...
- WithinReason 2y agoI'm not talking about mathematically exact reconstruction, I'm talking about not being able to see the difference.
- karmakaze 2y agoI first noticed it around 2014 on a 1080p Dell display (maybe ~24") with an i7 MacBook Pro of the day. It was the poor non-crisp pixels that led me to even knowing there were even RGB/YPbPr modes.
- WithinReason 2y agowas it maybe with 4:2:2 or 4:2:0 subsampling? at 4:4:4 there should be no difference