[FFmpeg-devel] [PATCH] QCELP decoder
Vitor Sessak
vitor1001
Fri Oct 3 23:06:26 CEST 2008
On Fri, Oct 3, 2008 at 6:52 PM, Kenan Gillet <kenan.gillet at gmail.com> wrote:
> hi,
> M, Kenan Gillet <kenan.gillet at gmail.com> wrote:
>> Hi,
>> On Oct 2, 2008, at 2:25 PM, Vitor Sessak wrote:
>>
>> For codecs that were supposed to be converted to signed 16-bit at some
>> point should be indeed between -1. and 1.. A good way to test if did not
>> do something wrong is to decode a sample to wav and see if your change
>> changes the result. It should either don't change or change a negligible
>> amount (see tests/tiny_psnr for evaluating how much it has changed).
>>
>> From the samples,
>> 101.mov
>> codeblue-interrupt01.mov
>> codeblue-interrupt06.mov
>> codeblue-plot08motor.mov
>> installfect_export.mov
>> oldepisode.mov
>> schlangenmannliten.mov
>> surge-1-16-B-Qclp.mov
>> 8fe3b9dcd6d7e6db3375521c5c1b_Video_041606_002.3g2
>> Test3gpp-mp4a.3g2
>> c8wwz2m0.3g2
>> qtaudio-qcelp-problem.3g2
>> test.3g2
>> tube.3g2
>> zg3dx2d6.3g2
>> vidoo_MP4_audio_Qcelp13k.k3g
>> tiny_psnr gives:
>> stddev: 0.00 PSNR:999.99 bytes: 250284/ 250284
Nice, this output says that the output didn't changed a single bit.
>> For blue_earth.mov,tiny_psnr gives:
>> stddev: 119.57 PSNR: 6.57 bytes: 19681324/ 19681324
>> For h263.mov, tiny_psnr gives:
>> stddev: 73.51 PSNR: 10.79 bytes: 13511404/ 13511404
>> is it still ok, or is something wrong?
> of course, something was wrong.
> After corrected a bug,
> the new values for this 2 files are:
> blue_earth.mov: stddev: 2.51 PSNR: 40.11 bytes: 19681324/ 19681324
> h263.mov : stddev: 1.81 PSNR: 42.95 bytes: 13511404/ 13511404
It is a pretty small difference, but I find it surprising that only
these two files changed. If you decode it twice with the same ffmpeg
binary, do you get the same output? If not, you could be reading some
uninitialized memory.
> is it now acceptable?
For me (and probably for the others), yes.
-Vitor
PS: it is always nice to see people finishing old soc projects, nice work!
More information about the ffmpeg-devel
mailing list