TMPGEnc: Deinterlacing PAL animation DVDs of US shows

[ralfbeckers]

Hi,
This post is interesting for all people in the European RC2 region.

Once in a while, we will come over a RC2 PAL disc that was sourced from an NTSC master. This can lead to having two different fields from the original NTSC master in ONE PAL field. The DVD looks pretty ugly then.

I bought South Park Volume 11 from England and looked at each field individually with TMPGEnc. There are about 20 true pictures/drwawings/frames for every 50 fields of video, but they are spreadf out over those 50 fields in the most unusual, irregular cadence. Every third frame of animation is showing in the same field as its successor. Sometimes there is not single field that shows the frame alone!

When you make a plain standard PAL (S)VCD from this DVD the results are awful.

Now I'm considering using the deinterlacing filter in a "off field, animation adaption" setting.

Who of you has ever done that? I'd like to start a very technical discussion which be base on our own experiments rather than opinion or theory. Anyone?

Regards
Ralf

[ralfbeckers]

Hi Shochan,

Love your documentation of the TMPGEnc setting in the Video Articles section of this site!

Could you enlarge in the Deinterlace setting in the Filters section in the future?

I believe it is like this: All SVCDs and DVDs store interlaced (two separate fields), even if you encode progressive (they just receive the respective flag). Of course, when you encode progressive, the fields A and B can be put together without artefacts because they are exact halves of the SAME frame. All (S)VCD also store at the field rate you encode to, no matter what the analogue output format is supposed to be. During the generation of the analogue output video the PLAYER decides what to do with a 24fps (=frames per second) disc.

VCDs, to my knowledge, store only one field, not two fields. That's why their spatial resoltuin is limited to half the lines of PAL or NTSC, respectively. The question is WHAT this single field (which is doubled during analog output) is made of, i.e. what goes in it.

Now, deinterlacing gives you the option to control how the above described A and B fields are derived. For example, yesterday I encoded an adult PAL RC2 disc to PAL VCD which was sourced from an interlaced video source such as BetacamSP. One second of ripped video has 50 fields, with all 50 fields referring to DIFFERENT points in time, no field is the same as the other. IF these where put on VCD without deinterlacing the playback on a TV with the usual interlaced scanning would maintain the interlace artefacts of the DVD plus adding motion noise because the A and B fields from the VCD are not exact clones of the A and B fields of the DVD stream because the spatial resolution is changed. Open such an interlaced source video stream in the deinterlace filter window and look at a few consecutive pictures which I believe to be plain composites of two fields already. With the deinterlacing filter disabled you see the interlace artefacts on all picture elements that have even the slightest bit of movement. Now scroll through the available filter settings. With an interlaced source such as the one described above the double (flied) setting works best. This seems to make "weave" the two A B source fields into one composite progressive field as good as possible, thus creating a progressive frame with maximum picture information/sharpness. During encoding the picture is sliced into two fields again but only now they belong to the same point in time. Technically speaking the resolution of time halves from 50 instances per second to 25 instances which should be OK for porn.

I also found the deinterlacing filter useful in burning a PAL VCD from a PAL DVD of South Park Volume 11. The British did make their PAL DVD from an interlaced NTSC source with has caused all kinds of weird problems. Sometimes it is impossible to find a frame that contains a given animation picture once and only once. What I mean is that animation such as South Park has e.g. 20 drawings per second which are mapped to 50 or 60 interlaced fields for the master video tape. When this is done, some pairs of A B fields on the master tape will already contain two different drawings! When such a tape is used as the source for making a DVD master tape of a different scan rate (60 --> 50 Hertz as in my case) the problem get MUCH more severe and the individual drawings can never again be separated from each other. This is where the field (animation adaptation 2) filter kicks in! Without the filter the VCD playback on a TV looks just as awful as the DVD because it will sometimes show a composite frame with three (!) different drawing elements in it!!! When using the filter this gets evened out.

I seriously suggest, we do a number of test runs with different sources with this kind of filter.
I assume that we test and master for the following application: Playback with a stand-alone VCD or DVD player, using its interlaced outputs to feed a TV that has no internal deinterlacing or 100 Hertz technology. This is the setting that 85 percent of us will have at home. Actually, it would be interesting to learn what people with progressive scan DVD players experience when they play VCDs which were properly deinterlaced before mastering.

Mit freundlichem Gruß / Kind Regards

Ralf Beckers


<font size=-1>[ This Message was edited by: ralfbeckers on 2001-09-21 08:56:18 ]</font>