2009/1/19 Stas Oskin <[email protected]>: > Correct. But the GOP size and bit rate shouldn't really be considered >> as being so dependent. Bit rate is generally an independent way to >> increase the quality. >> >> Think of it this way - the options I gave you before make the encoder >> work harder to increase the quality per bit of the compressed stream. >> Usually one has to consider a speed/quality trade off for one's own >> situation. The bit rate should be adjusted to your needs depending on >> your playback device and available storage/bandwidth usually. > > > Thanks, now finally I understand where the GOP fits. > > I was advised some time ago (by Luca, if I'm not mistaken) about the > following options, which should improve the quality from the start: > > c.mb_decision=2;//FF_MB_DECISION_RD > c.noise_reduction=250; > c.pre_me=0; > c.me_method= 5; > > I think the mb_decision relates to the -mbd 2 flag you told before about. Do > you know about the other options as well?
Correct. Noise reduction is probably the one that's done during encoding. Less noise allows for better compression but too much noise reduction will make the image look soft and remove details. pre_me = 0 disables pre motion estimation which I think is some sort of motion estimation (searching for motion vectors) done before the full motion estimation to obtain more information to make better predictions. I would have thought enabling it would aid compression. me_method = 5 currently corresponds to the diamond motion estimation pattern but I think these have changed in the past few months so maybe it previously corresponded to something else. The options I suggested should be fine quality wise. Can you give more information about what your target is like (frame rate, resolution, bit rate)? Regards, Rob _______________________________________________ libav-user mailing list [email protected] https://lists.mplayerhq.hu/mailman/listinfo/libav-user
