[Mp4-tech] About calculation of SAD in JM10.2

[Tourapis, Alexis]

In the updated (not yet released) reference software we have completely
rehashed the motion estimation process. The code makes it way simpler to
add your own distortion model (currently it supports SAD, MSE, and SATD
for all possible refinements). Chroma can also be optionally used for
distortion computation as well. You may be able to contact Karsten
Suehring if you would like to get a hold of the software before its
release.
Best regards,
Alexis
  _____  
From: mp4-tech-bounces lists.mpegif.org
[mailto:mp4-tech-bounces lists.mpegif.org] On Behalf Of Andrew Krupiczka
Sent: Tuesday, November 07, 2023 9:08 AM
To: A R; mp4-tech lists.mpegif.org
Subject: RE: [Mp4-tech] About calculation of SAD in JM10.2
Amir,
I feel like you might have some misconceptions here.
Typically, you want to know SAD on block basis i.e. you compute pixels
difference (byte_abs[ *orig_line++ - *ref_line++ ];) from two different
frames for multiple pixels e.g. rectangular block of pixels. In JM
software, you should be able to find a multiple for(;;) loops iterating
through multiple pixels positions in two different frames (pointed to as
in the example below by ref_line and orig_line pointers). SAD values as
in the example below are cumulated on top of initial motion cost mcost
value if you want to get a better trade-off between a target bitrate and
video quality, but you may want to have a separate routine for
computation of SAD (and eventually modifying) like in the SATD();
    for (y=0; y<blocksize_y; y++)
    {
      ref_line  = get_ref_line (blocksize_x, ref_pic, cand_y+y, cand_x,
img_height, img_width);
      orig_line = orig_pic [y];
      for (x4=0; x4<blocksize_x4; x4++)
      {
        mcost += byte_abs[ *orig_line++ - *ref_line++ ];
        mcost += byte_abs[ *orig_line++ - *ref_line++ ];
        mcost += byte_abs[ *orig_line++ - *ref_line++ ];
        mcost += byte_abs[ *orig_line++ - *ref_line++ ];
        ...
      }
Hope it helps. Best regards,
Andrew
-----Original Message-----
From: mp4-tech-bounces lists.mpegif.org
[mailto:mp4-tech-bounces lists.mpegif.org] On Behalf Of A R
Sent: Tuesday, November 07, 2023 6:47 AM
To: mp4-tech lists.mpegif.org
Cc: sundarhills hotmail.com
Subject: [Mp4-tech] About calculation of SAD in JM10.2
Hi dear researchers !
I am working on H.264's reference model JM10.2 . In this reference code
I want to change the SAD with my own method. As a big picture I know
that, in SAD, two image values(as parameters) are required in
calculating SAD between them. The problems I am facing right now are:
1- I could not find/identify these two parameters in this reference
code.
2- I could not find/Identify the methods or functions in the code which
are dedicated to calculate the SAD.
Note: I found three functions/Methods namely SATD() , SATD8X8() and
find_SAT() but they do not answer any of the questions above.
I'll really be grateful if someone can help in this regard. Thanks.
Kind Regards,
Amir Raza. (LUMS, Lahore, Pakistan)
  _____  
  _____  
Try the next generation of search with Windows LiveT Search today! Try
it! <http://search.live.com> 
-----------------------------------------
This message (including any attachments) may contain confidential
information intended for a specific individual and purpose.  If you
are not the intended recipient, delete this message.  If you are
not the intended recipient, disclosing, copying, distributing, or
taking any action based on this message is strictly prohibited.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: /pipermail/mp4-tech/attachments/20061107/65a818d8/attachment-0001.html