On Wed, Sep 17, 2003 at 07:57:14PM -0700, Steven M. Schultz wrote: > > On Wed, 17 Sep 2003, Richard Ellis wrote: > > > is driven by Q (in that section at least) was reworked quite a bit in > > January. That would account for the difference I'm seeing in the new > > mpeg2enc's Q that does not seem to be having much effect at times. > > At least not with the default value of -X (100). I think > experimenting with -X could make -Q have more effect.
Increasing -X a little (200 vs. 100) does seem to make -Q have a bit more effect, if the fact that the quant=xx.xx value bounces around much more during encoding is partially caused by -Q. > > > Perhaps -N <num> (num =0.0 to 2.0) would work as well. In > > > 1.6.1 > > > That is a good suggestion, thanks. The new adjustable -N value > > should be able to accomplish much the same level of "fine tuning" > > That's what I've been using it for. The higher the value the > smaller the file (because the high frequency quantizers are raised > - high -N values will lose detail). -N 0.5 is very mild, 1.0 I've > found to be a good middle point for preserving detail but still > saving bits. The -N parameter does work quite well for shrinking a file, but it seems it's a bit sensitive. In a test run, with -N 0.0 I got a file size of 748,032 kbyte. With -N of 0.1 I got 708,956 kbyte on the same input. The 40meg loss did drop it right into my target size window, so it worked for that, but I was expecting less drop from only 0.1. > So far I've settled on '-K kvcd' as producing very good pictures > (the dark scenes look better than before) but perhaps the tmpegenc > tables might, on some material look better. I did try the kvcd matrices on a couple tests, and they did seem to produce very good images with a low bit rate as well. But upon close inspection of the TV output, it also seemed like I could see some edge ringing on sharp contrast changes. ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Mjpeg-users mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/mjpeg-users
