[Mp4-tech] [H.264 AVC/ Quantization]

[Gary Sullivan]

Ms. Bacquet et al,
I suggest getting a more recent copy of the standard.  It includes support for frequency weighting matrices, in the High Profile (which is the primary profile used for high-quality applications) and in some other profiles.  Only the very old orginal version did not support that.
In the first version, the transform block size was so small that there might not have been a benefit.
Also, when considering the tricks that can be put into an encoder, that feature may not be as necessary as was orginally believed in MPEG-2 days. (However, we ultimately did consider it desirable to support it in the High profile, and I'm glad that we did.)
Best Regards,
Gary Sullivan
________________________________
From: mp4-tech-bounces lists.mpegif.org [mp4-tech-bounces lists.mpegif.org] On Behalf Of Anne Sophie Bacquet [annesophie.bacquet gmail.com]
Sent: Tuesday, May 13, 2024 11:56 PM
To: mp4-tech lists.mpegif.org; xavier coudoux
Subject: [Mp4-tech] [H.264 AVC/ Quantization]
Dear Experts,
I have a question about H.264/AVC quantization.
In the previous MPEG.2 compression standard, the quantization is psychovisually weighted: the quantization step values for Intra frames vary according to frequency range.
Because SVH sensitivity is higher for low and middle spatial frequencies, the corresponding step values are lower.
But now, in the H.264/AVC case, the quantization is uniform. So, what about psychovisual weigthing and SVH sensitivity ? Does anyone have the explanation ?
Is there any specific document where I can find the theoritical justification of AVC uniform quantization ?
Thanks in advance.
Best regards,
Anne Sophie Bacquet
IEMN- Valenciennes
-------------- next part --------------
An HTML attachment was scrubbed...
URL: /pipermail/mp4-tech/attachments/20080514/1679cfae/attachment.html