[M4IF Technotes] Distortion Testing for Encoder

[Rob Koenen]

Please allow me write a Word of Caution.
While 'objective' measures may be starting to get better than SNR, 
and are highly useful for monitoring purposes and optimizing encoder 
quality (including making coding decisions), they do not allow one 
to choose between one coder and the other. 
Their correlation with human subjective quality judgement is still too 
low for such purposes.
Kind Regards,
Rob
-----Original Message-----
From: Jeyendran Balakrishnan [mailto:jp   skystream.com]
Sent: Wednesday, August 14, 2023 14:36
To: 'manu.batura   patni.com'; technotes   lists.m4if.org
Subject: RE: [M4IF Technotes] Distortion Testing for Encoder
Try the JND Toolkit from David Sarnoff labs 
( http://www.sarnoff.com/internet_telecom/jndmetrix/index.asp
<http://www.sarnoff.com/internet_telecom/jndmetrix/index.asp> ) 
available for about US$ 9K, Windows SW, offline analysis. 
Or you can get the more pricey Tektronix PQA box (about $30K 
I think, but don't hold me to this), which uses the same 
Sarnoff code inside, but (I believe) works real-time and 
for streaming content, and probably adds other bells 
and whistles. 
This generally gives some objective measures of video quality, 
taking into account the nature of the human psycho-visual system. 
You can use this to compare quality of different encoders 
on the same input content. 
However, to decide if any one encoder is good enough on 
an absolute basis is subjective, and generally 
addressable only by having a (sufficiently large) 
bunch of people look at the reconstructed output and judge 
the quality. 
-Jeyendran 
-----Original Message----- 
From: Manu V Batura [ mailto:manu.batura   patni.com
<mailto:manu.batura   patni.com> ] 
Sent: Tuesday, August 13, 2023 11:24 PM 
To: technotes   lists.m4if.org 
Subject: RE: [M4IF Technotes] Distortion Testing for Encoder 
Thanks Rob, 
But are there any objective tests for distortion that can be used 
for testing which encoders are better. If distortion is the competitive 
measure, there 
should be distortion thresholds available that may be used to call some 
encoders good 
and some bad. 
Thanks & Regards, 
Manu 
-----Original Message----- 
From: technotes-admin   lists.m4if.org 
[ mailto:technotes-admin   lists.m4if.org
<mailto:technotes-admin   lists.m4if.org> ]On Behalf Of Rob Koenen 
Sent: Wednesday, August 14, 2023 10:54 AM 
To: 'manu.batura   patni.com'; technotes   lists.m4if.org 
Subject: RE: [M4IF Technotes] Distortion Testing for Encoder 
Indeed, any decoder that produces a valid bistream is compliant 
to the standard. There are deliberately no prescribed quality 
metrics, nor is there any prescribed algorithm to perform the 
encoding. 
This means there will be good encoders and bad ones... 
Best, 
Rob 
> -----Original Message----- 
> From: Manu V Batura [ mailto:manu.batura   patni.com
<mailto:manu.batura   patni.com> ] 
> Sent: Tuesday, August 13, 2023 20:53 
> To: technotes   lists.m4if.org 
> Subject: [M4IF Technotes] Distortion Testing for Encoder 
> 
> 
> Hi, 
> 
> As per Part-4 of the standard, it seems that any video 
> encoder that can 
> generate a compliant bitstream would be 
> compliant to the standard. However, is there aso a 
> corresponding distortion 
> measure that needs to be acheived? 
> (something like RMSE etc). If yes, what are the thresholds 
> that should be 
> used for testing 'acceptable video quality'/'allowable video 
> distortion'? 
> 
> Thanks & Regards, 
> Manu V Batura 
> 
> _______________________________________________ 
> Technotes mailing list 
> Technotes   lists.m4if.org 
> http://lists.m4if.org/mailman/listinfo/technotes
<http://lists.m4if.org/mailman/listinfo/technotes>  
> 
_______________________________________________ 
Technotes mailing list 
Technotes   lists.m4if.org 
http://lists.m4if.org/mailman/listinfo/technotes
<http://lists.m4if.org/mailman/listinfo/technotes>  
_______________________________________________ 
Technotes mailing list 
Technotes   lists.m4if.org 
http://lists.m4if.org/mailman/listinfo/technotes
<http://lists.m4if.org/mailman/listinfo/technotes>  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: /pipermail/mp4-tech/attachments/20020814/8275cdd0/attachment.html