evilclive
29th December 2006, 19:58
I would like to use DivX Pro 6.4 to compress my huge collection of MPEG-1 files, to save some hard disk space.
The files were recorded using an original All in Wonder Radeon and MMC 7.7 software, from a PAL source with very little noise or ghosting. I used 2400 kbit/s at 352x288 pixels for the video - a good 70% higher than the recommended bit rate, and the files contain occasional long strings of 0xFF bytes, where there wasn't enough video data in the uncompressed source to fill the CBR stream.
I have tried transcoding them using VirtualDubMod and the DivX codec, with a one-pass quantiser-based setting, light noise reduction and the Balanced speed/quality preset. Turning on B-frames makes the files about 10% smaller for a given quantiser setting.
At a quantiser of 3, the transcoded files are scarcely any smaller than the source files, which defeats the purpose of transcoding them. A quantiser of 5 gets the DivX video files down to about 1000 kbit/s, which is what I want, but I know the usual recommendation given on this forum is to set the quantiser to 2, 2.5 or 3 when space is not that critical, and to use a two-pass encoding otherwise.
As MPEG-4 is meant to be a more efficient codec than MPEG-1, if my target files are the same size as the source files, all the extra data must be encoding artifacts in the source data, rather than the original signal.
I don't think I'm doing anything wrong, other than perhaps being a bit optimistic in trying to halve the size of all my files, without appreciable loss of video quality, just by using a more efficient codec.
Can I really hope to get near-DVD quality, albeit at quarter resolution, at around 1.0 Mbit/s video plus 160 kbit/s (ABR MP3) audio?
Is the art of DivX compression to obtain a bit rate you desire for a quality you're prepared to tolerate, rather than the other way round?
The files were recorded using an original All in Wonder Radeon and MMC 7.7 software, from a PAL source with very little noise or ghosting. I used 2400 kbit/s at 352x288 pixels for the video - a good 70% higher than the recommended bit rate, and the files contain occasional long strings of 0xFF bytes, where there wasn't enough video data in the uncompressed source to fill the CBR stream.
I have tried transcoding them using VirtualDubMod and the DivX codec, with a one-pass quantiser-based setting, light noise reduction and the Balanced speed/quality preset. Turning on B-frames makes the files about 10% smaller for a given quantiser setting.
At a quantiser of 3, the transcoded files are scarcely any smaller than the source files, which defeats the purpose of transcoding them. A quantiser of 5 gets the DivX video files down to about 1000 kbit/s, which is what I want, but I know the usual recommendation given on this forum is to set the quantiser to 2, 2.5 or 3 when space is not that critical, and to use a two-pass encoding otherwise.
As MPEG-4 is meant to be a more efficient codec than MPEG-1, if my target files are the same size as the source files, all the extra data must be encoding artifacts in the source data, rather than the original signal.
I don't think I'm doing anything wrong, other than perhaps being a bit optimistic in trying to halve the size of all my files, without appreciable loss of video quality, just by using a more efficient codec.
Can I really hope to get near-DVD quality, albeit at quarter resolution, at around 1.0 Mbit/s video plus 160 kbit/s (ABR MP3) audio?
Is the art of DivX compression to obtain a bit rate you desire for a quality you're prepared to tolerate, rather than the other way round?