How can I fix VOB's with incorrectly interleaved fields?

[LisaB]

I have a VOB that has a problem.

When the video stream is viewed in VirtualDub, it appears that every frame is interlaced. I have discovered that what is actually happening is that the fields have been incorrectly combined into frames. In other words, the top field of one frame matches (temporally) the bottom field in the following frame.

If I use the following AviSynth script, then the output appears perfectly (i.e., nice perfectly progressive frames).

MPEG2Source("k:/sampvideo.d2v")
SeparateFields()
DeleteFrame(0)
Weave()

All this script does is separate the frames into fields, delete the first field, and then recombine the fields back into frames.

I have no idea how to fix this. I don't want to re-encode, because it shouldn't be necessary. I just want to fix the interleaving of the fields so I get nice progressive frames out...

Lisa B.

[FOO]

Sorry , you must re-encode. If it was just every pair of
fields swapped , you could fix it with Restream.
That script is operating on uncompressed video and produces
uncompressed video.

Do it overnight and it won't hurt so much

edit: note that what you have described makes the matching fields
4 fields apart - 1T/1B 2T/2B if it's top field first.
See if Restream reports bottom first

[LisaB]

Hi FOO,

Well, I must have described it wrong, then....in fact the *bottom* field of one frame matches temporally with the top field of the next. In any event, the two fields that should be together are right next to each other, it's just that they have been "packaged" into different frames. If this weren't the case, then my avisynth script wouldn't have fixed it.

Now, when you encode something in TMPGEnc, for example, there are two encode modes - progressive and interlace. I was always under the assumption that if you encoded using interlace mode, then the fields would be encoded as distinct units. In the case of a truly interlaced source, this would lead to a much more efficient and high-quality encode.

Imagine, for instance, creating an image in PhotoShop that is composed of alternating red and green horizontal lines. If you save as a jpeg, it tries to compress this image with all these sharp lines, and the result isn't so great. However, if you separate it into fields, then you would have a solid red field and a solid green field - imagine how easy these would be to encode the two fields separately as distinct images! Anyway, this is how I imagine encoding using interlace mode, and why I feel that it is more efficient for true interlaced sources. If you're encoding progressive frames, then it probably doesn't matter quite as much which mode you use...

I've never seen anywhere a reasonable description of encode modes, so this is all just speculation.

Anyway, assuming (hoping) that my VOB was encoded using interlace mode, then I figured I could somehow just delete the first field, and then essentially "re-label" the fields (i.e., 1T,1B,2T,2B...) so that they are correct.

Also, what confuses me even more is that this VOB, when burned to DVD, plays beautifully on my TV. When stepping through frame by frame on the standalone DVD there is no interlace! But when I try the same thing using a software DVD player on the computer - interlace! And, before you point out that TV's are field-based, I would add that I have an HDTV, and that I was using a progressive-scan DVD player in progressive mode!

Last question - is this the right forum for this question, or should it be in "Advanced Conversion"

[FOO]

MPEG2 has "pictures" which are frames.
they are either stored as 1 frame or 2 sequential fields.
The MpEG2 stream has flags that control the display order
they can be changed without re-encoding. This has the effect of
swapping every pair of fields. The MPEG spec Requires that
field based pictures come in pairs , so you can't delete just one field.

Besides , if it plays OK why are you complaining ?

I know of no programs that can restructure field based MPEG2
even though it might be possible without re-encoding