Capture 'real' frames
Posted: Mon Apr 11, 2011 4:59 am
I wish to do some experiments to compare the quality of different encoders and the effect of B-frames when converting to MPEG2/MPEG2-PS and MPEG4/AVI Progressive.
To do this, I thought to frame step each clip in SMplayer 0.6.9 and compare the frames. Preferably, by capturing critical frames.
I did captures in Pause mode in VirtualDubMod for a bitrate of 1000 kbps and there were differences between B = 0 and 2, but the overall difference in quality was not significant.
The variation within each clip was as great as the change due to changing from 0 B-frames to 2. This finding is counter intuitive since B-frames are supposed to make more efficient use of the available bit rate! The conversions with 2 B-frames took 50% longer than with none.
On the other hand, early results for SMplayer suggest a different conclusion, using the “Stop/Start taking screen-shots” mode.
All the B- and P-frames for the file with B set at 2 looked better than for B=0! This is what one would expect. It could be, then, that SMplayer corrects any B- and P-frames using data from adjacent I-frames, as a player should.
However, this explanation fails to explain why the I-frames at both values of B were smoother, with fewer artefacts, than the same frames viewed in VirtualDubMod.
This is contrary to the fact that I-frames should not change when played (no correction).
Hence, maybe the B- and P-frames are better in SMplayer because it uses a smoothing technique?
Please: any comments on the 'reality' of frames captured by SMplayer?
I want the captured frames to be a realistic means of indicating video quality!
To do this, I thought to frame step each clip in SMplayer 0.6.9 and compare the frames. Preferably, by capturing critical frames.
I did captures in Pause mode in VirtualDubMod for a bitrate of 1000 kbps and there were differences between B = 0 and 2, but the overall difference in quality was not significant.
The variation within each clip was as great as the change due to changing from 0 B-frames to 2. This finding is counter intuitive since B-frames are supposed to make more efficient use of the available bit rate! The conversions with 2 B-frames took 50% longer than with none.
On the other hand, early results for SMplayer suggest a different conclusion, using the “Stop/Start taking screen-shots” mode.
All the B- and P-frames for the file with B set at 2 looked better than for B=0! This is what one would expect. It could be, then, that SMplayer corrects any B- and P-frames using data from adjacent I-frames, as a player should.
However, this explanation fails to explain why the I-frames at both values of B were smoother, with fewer artefacts, than the same frames viewed in VirtualDubMod.
This is contrary to the fact that I-frames should not change when played (no correction).
Hence, maybe the B- and P-frames are better in SMplayer because it uses a smoothing technique?
Please: any comments on the 'reality' of frames captured by SMplayer?
I want the captured frames to be a realistic means of indicating video quality!