[Mp4-tech] [H.264]about the scaled versions of the motion vector in the 8.4.1.2.3 section

[JK Kim]

Hi Tan,
If all the roundings and clippings are omitted and a substation is made,
You can simply get the typical linear interpolation equation.
tx = ( 16 384 + Abs( td / 2 ) ) / td
(8-194)
    = 2^14 / td
DistScaleFactor = Clip3( -1024, 1023, ( tb * tx + 32 ) >> 6 )
(8-195)
                         = 2^8 * tb / td
mvL0 = ( DistScaleFactor * mvCol + 128 ) >> 8
(8-196)
          = (2^8 * tb / td * mvCol) >> 8
          = (tb / td) * mvCol
mvL1 = mvL0 - mvCol
(8-197)
          = (tb / td) * mvCol - mvCol
          = ((tb-td) / td) * mvCol
IMHO, (8-194)-(8-197) is designed to prevent overflow and reduce the number
of division.
Experts, please let me know if more!
                  JK
  _____  
From: mp4-tech-bounces lists.mpegif.org
[mailto:mp4-tech-bounces lists.mpegif.org] On Behalf Of Tan Rui -- Huawei
Sent: Monday, August 07, 2023 7:16 PM
To: mpeg 4 mail-list
Subject: [Mp4-tech] [H.264]about the scaled versions of the motion vector in
the 8.4.1.2.3 section
Hi, All experts,
I want to ask a question about the calculation about the scaled versions of
the motion vector in the 8.4.1.2.3 section.
I read lots of papers give the following equations:
mVL0 = (tb / td) * mvCol
mVL1 = ((tb - td) / td) * mvCol
But in the protocol 8.4.1.2.3 section, from the (8-194) equation to the
(8-199) equation, there is a DistScaleFactor 
variable. I can not calculate those equations from the above equations which
were found the papers. Can
some one provide some information about it? I spent some time to search the
detail description, but I can not
get any useful explanations.
Tan Rui
2023/8/8
-------------- next part --------------
An HTML attachment was scrubbed...
URL: /pipermail/mp4-tech/attachments/20060808/9d122ba5/attachment-0001.html