[M4IF Technotes] Re: Codec testing methodology

Ralph Sperschneider sps iis.fhg.de
Mon Jan 21 16:31:48 EST 2002


Dear Gautam, all,
Gautam Kumar wrote:
> 
> Dear all,
>  I have some doubts regarding comformance testing.
> IS it sufficient for a decoder to pass ISO confermance test?
It is necessary, but most probably not sufficient. The tests are defined in a way that they check most of the features of most of the tools, but by no means any particularly possible combination.
> If a decoder passes all the ISO conformance test streams then it simply
> means that
> varible length decoding along with some of (not all) the paramerters are
> decoded correctly.
> What i mean to ask is that how can we know that the quality and
> the information in the decoded stream is the actual desired stream ?

By comparing it with the output of the reference decoder or with the provided reference sequence, using the comparison tests specified by MPEG in the conformance part of the standard.
> Should not there be the desired decoded stream or the original streams
> those have been used for encoding along with the confermance test streams ?
Yes, the desired decoded stream should be usually there (there might be some special sequences, where other data is provided instead of the decoded reference). A comparison with the original makes not much sence.
> So that we can compair our decoded stream bit by bit with the intended one.
A bit by bit comparision is most times not possible due to rounding issues. But at least for Audio there are comparison tools available which perform the required tests an deliver a PASS/FAILED output.
> Besides that how can one test his MPEG4 Video ENCODER ?
> i can guess only about the interoperability test.
> Is there any other way to test Encoder ?
Wrt to conformance, you might take any available decoder (preferably the reference decoder), and you might want to add some checks to test the individual data elements regarding theyr conformance criteria as defined in the conformance part of the standard. Since you only need the parser, you might even consider to remove the actual decoding part from the software, if speed counts. Testing the quality of the encoder is another issue, but most likely out of scope here.
Best regards,
Ralph
-- 
Dipl.-Ing. Ralph Sperschneider  | Phone: +49 9131 776 344
FhG IIS-A                       | Fax:   +49 9131 776 398
Am Weichselgarten 3             | mailto:sps   iis.fhg.de 
D 91058 Erlangen                | http://www.iis.fhg.de/amm/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: sps.vcf
Type: text/x-vcard
Size: 336 bytes
Desc: Card for Ralph Sperschneider
Url : /pipermail/mp4-tech/attachments/20020121/aca4085f/sps.bin


More information about the Mp4-tech mailing list