Different flavors of chroma sampling at 4:2:0

[shae]

Is there somewhere an exhaustive explanation of the various chroma sampling methods used for 4:2:0 (and others like 4:2:2, etc. if there are any nuances there)?

I just found out there are differences in how chroma is sampled at 4:2:0 between MPEG2, MPEG1, DV, etc.

Would that imply storing, for example, DV as Lagarith YV12 might lead to a loss?

[JohnnyMalaria]

They differ in their horizontal and vertical co-siting with the luma. In 4:2:0 DV, Cb and Cr alternate line by line. In MPEG-2, Cb and Cr are cosited horizontally.

See the images on:

http://msdn.microsoft.com/en-us/library/ms694252(VS.85).aspx

So, some loss will occur converting to/from DV 4:2:0 and MPEG-2 4:2:0. But this will be no where near as much as DV 4:1:1 and MPEG-2 4:2:0.

[shae]

Thanks.

So in the case of MPEG2 it drops half the horizontal chroma while JPEGs averages also horizontally?

Is Vertically_AlignedChromaPlanes only used for interlaced?

Can you suggest anyplace showing image comparisons of source images vs. the various samplings methods?

Are there only 3 sampling variants?

What does MPEG4 use?

[mpucoder]

JPEG can sub-sample chrominance horizontally, vertically, neither, or both.

[shae]

Okay. But I guess whenever it does it picks the chroma from the same locations which are the average of neighboring pixels.