[M4IF Technotes] Chroma components in inter4v mode

[Gary Sullivan]

+> -----Original Message-----
+> From: Christoph Lampert [mailto:chl   math.uni-bonn.de] 
+> Sent: Monday, June 24, 2023 3:42 AM
+> To: technotes   lists.m4if.org
+> Subject: RE: [M4IF Technotes] Chroma components in inter4v mode
+> 
+> 
+> So the answer is: "It was chosen that way, because it does 
+> not matter very much". Okay, that's an answer. 
Right.
+> How's the situation in H26.L? 
Now that question's more fun.  In the new codec
(subclauses 9.2.1.4 and 9.2.2.3 of the Joint Committee Draft):
1) The number of motion vectors for chroma is the same
as the number of motion vectors for luma (you probably
like that).
2) If the luma block size is MxN, the corresponding
chroma block size is (M/2)x(N/2).  This is to compensate
for the difference in resolution between luma and chroma.
3) If the luma motion vectors have 1/4-sample resolution,
the chroma MVs have 1/8-sample resolution (again
compensating for the difference in resolution).
4) Correspondingly, if the luma MVs have 1/8-sample
resolution, the chroma MVs have 1/16-sample resolution.
5) Although the interpolation for luma uses some
longer filters, chroma always uses bilinear
interpolation.
So you will probably like the new design better, although
the sloppier interpolation filters (item 5) still reflect
some degree of reduction in the priority given to chroma
relative to luma.
Best Regards,
-Gary