[M4IF Technotes] how to quantize intra_dc using H.263 method?

[Gary Sullivan]

Actually, there are an infinite number of ways to quantize.
It is a popular misconception that video coding standards
tell you how to quantize data -- they don't.  Even some
academic publications encourage this serious misconception.
The standards actually let you quantize any way you want.
In fact I don't think they even require the encoder to quantize
anything at all.  What they do specify is how the decoder will
interpret the syntax it is sent.  It is entirely up to the encoder
designer to figure out how to send it something useful.
Yes, there is syntax in the MPEG-4 standard to tell the
decoder how to interpret the coefficient indices.  There are
basically two different inverse quantization methods
supported in MPEG-4 video.  In MPEG-4, there is one called
"type 1" and another called "type 2" (section 7.4.4).
Type 1 inverse quantization is similar to that in H.263 baseline
and Type 2 inverse quantization is similar to that in MPEG-2.
However, I don't think either one is exactly the same as in
those other standards -- except for when the "short header"
format is used, in which case it exactly follows H.263 baseline.
Best Regards,
Gary Sullivan
+> -----Original Message-----
+> From: wuks   bhnec.nec.co.jp [mailto:wuks   bhnec.nec.co.jp] 
+> Sent: Monday, April 15, 2024 8:15 PM
+> To: technotes   lists.m4if.org
+> Subject: [M4IF Technotes] how to quantize intra_dc using 
+> H.263 method?
+> 
+> 
+> when using H.263 method to quantize intra MB's DC, there are 
+> two means to
+> use,
+>   one is level=cof//8;
+>   the other is as same as MPEG's;
+> 
+> but how does the decoder dequantize( or is there a flag to 
+> differentiate
+> them) ?
+> 
+> 
+> 
+> _______________________________________________
+> Technotes mailing list
+> Technotes   lists.m4if.org
+> http://lists.m4if.org/mailman/listinfo/technotes
+>