[RESOLVED] Glitch in decoding #2 - this time replicable

[Nick007]

I was lucky and discovered this glitch occuring right at the beginning of movie, frames 448-454. In the glitch, it shows blended frames from few hundreds frames back.
The glitch is there only if I frame step it in Avsp (fast with right arrow) or play it in mpc-hc/Virtual dub, the glitch is not there if I seek to the frames directly or seek to frame 200 for example and frame step from there.



25 sec sample cut with tsMuxer: http://www.mediafire.com/?500cww7ldl6zcy7

Windows 7 64-bit, 32-bit DGDecNV, Avisynth 2.5.8, Nvidia GTX 570 with Forceware 285.62.

This glitch seems to be very similar to ones I had some time ago and which I described in earlier topic, but this time I was able to reproduce it on my laptop which has GT 220 M. It may be caused by a glitch in the source, but it's strange that in this place is no glitch when I use dss2 and ffdShow for H.264 decoding and the glitch is not there when I seek directly to or anywhere before those frames, the frames are okay. It seems to me, as if some frame on the beginning (with the Warner Bros logo) caused corruption which manifested after few hundred frames. The appearance of the glitch is very similar to those I experienced earlier and from what I read also others.

Maybe if you find out what caused it, it could also resolve the other topic

[admin]

Sorry for the delay in getting to this.

It sure looks like a problem in your source stream, as I am able to see the same artifact in VLC player.

[Nick007]

Yeah, the glitch was somewhere in the source (it was there with CoreAVC too), but if I seek 100 frames before that frame 448 and then frame-step to 448, there's no glitch in output.
Only if I frame-step from frame 0, it causes the glitch in output. I wonder why is that?

[admin]

Could be that SPS/PPS are affected such that the starting point determines which SPS/PPS is active at the critical moment.

[admin]

Marking resolved as it is clearly a problem in the source stream.