I was reading that today... well it's a compromise. I used it especially to convert NTSC to PAL kvcd. I liked the results as seen on my tv..
Now with KDVD, well so far when I backup an NTSC dvd the
destination is an NTSC KDVD (playback pulldown) and I am not deinterlacing this source at all (I use forced film in these cases).
I was thinking whether it would have been worth it to deinterlace and convert to PAL KDVD. Probably not at all - it would take more time and
due to the increased PAL resolution less would fit on a dvd-r.

Recently I did some PAL KDVD encodes and I did deinterlace them, one was a PAL DV source. The result was excellent. I'll have to experiment next time without deinterlacing to see if I can get better results on kdvd.
I'm not sure about non-NTSC film sources, deinterlacing might be better for those, or perhaps not, this would be interesting to experiment and see the results.

Because the KVCD matrix is what makes the difference.
Any KDVD will be better (or at least equal) to any DVD, but at less bitrate which results in less file size.Compared to a regular DVD target, KDVD will always be better, be it progressive or interlaced.

-kwag

Regarding stuff of NTSC origin, then transferred to PAL:

If you have sources that were originally shot at 23.976fps, the PAL version should be progressive. If it's not, someone has screwed up something during the NTSC->PAL transfer, often resulting in a fieldblended (=ugly looking) material. In this case it might be useful to do a 25fps->24fps conversion with Restore24 and then do a 24fps->25fps conversion via AssumeFPS(25.000). I had to do this on a Citizen Kane version and my creation looks a lot better than the original (now who said that you can't get better results than the source is :lol: )

If the original material was shot at 29.97fps (mostly the stuff meant for TV, also most concert footage), the PAL version should be pure interlaced video. This means that deinterlacing it will result in less smooth motion, and you'll also lose sharpness and details!

Regarding stuff of PAL origin:

Most (all?) PAL movies (that is, material which is not created primarily for TV broadcasting) are progressive so no need to deinterlace.

Most material meant for TV broadcasting is pure interlaced video, which means that on TV the motion looks very smooth. If you deinterlace this material, you'll lose the smoothness along with some sharpness and details.

Then there's also the phase shifted stuff inc mentioned.

another Phil's phrases to my "hall of fame"!
8O :rotf:

:dunce:

I recorded some music videos from satellite (dvb-s card, 25fps pal interlace source) and encoded them in pal kdvd encoded format. Well the result was lots of ugly interlacing lines on my 100hz widescreen tv. I decided the best thing was to deinterlace using kerneldeint so I rencoded today and wrote another kdvd (non-interlaced however). The new kdvd looks better on my tv, no visible loss of sharpness or details. The other one looks plain terrible, with visible interlace lines. The music videos did have lots of motion in them, I might add, and this made it worse.

I had two avisynth scripts for this kdvd (2 different files), one which did nothing but resize and add black borders, and another one which did some filtering.

On my pc the interlaced dvd looks ok with media player classic's built in mpeg2 deinterlacer. But for my tv and/or dvd player the story seems different. I guess the best thing is to test and see what results you get with your equipment.

Then you've got something really wrong either in your setup or your source. Did you check that you had the correct field order set in your encoder? TV should never show combing!

Also you should never directly filter or resize interlaced sources, they must be treated in a special way (there are threads on the forum about that).

And thats the point!!

IF you capture (or stream via DVB) Musicvideos in PAL areas, 90% of these videos where produced in the US., means in NTSC area.

Now, those Music-Videos often are even shot on 24fps film (not concerts!), but they will be telecined to 29.97 where at THIS State the cutting/postprocessing will be done. So the interlacing pattern will change! Well if you would IVTC such captures at this 29.97 state, that would be no problem as you would end up easier in a 23.976 progressive restored music video.
BUT: These 29.97 Music Videos will be ported to the PAL market, and they wont do a IVTC before and then a speedup to 25.000, no! They do perform on that 29.97 source a direct norm-conversion to 25.000fps PAL interlaced!
THATS why 90% of all PAL captured US Videos will come out incl. blendings when performing a simple deinterlacing on such material.

The (IMHO) only "automatic" way to get US Music Videos on PAL DVDs or PAL captures back to their orig. progressive state is to treat them via Restore24().
But restore24() is still in developing phase and (for my personal taste) the treated streams still come out incl. some doubled Frames etc. as the blending recognision of restore24() is still not optimal. But anyway a GREAT funtion/approach by Didée.


And the story continues:

Last year 95% of all my movie captures came as 25.000fps progressive (speedup) on my HD. And I could directly convert them to mpeg1/mpeg2.
NOW the broadcasting stations (here in Germany) changed their way:
50% of all movies now do come out treaten by such fps conversion boxes!
They do use that way to compensate cutted adult scenes or if they want to "stretch" the movietime to get more allowed commercials into it. This will be compensated by using such horrible fps conversion techniques.

Like yesterday, when I tried to capture "HotShots". The result was a progressive/interlaced mixture, incl. pattern changes and so on.

And I (IMHO) do think, that "they" know that many users do capture and burn to DVD-R .... and thats also a way to make it harder for "us". So "they" maybe want us to inverst more in buying orig DVDs :lol:

My setup was correct (except for the interlaced resizing of course), it was the source basically as incredible said. I even recorded mostly from an fta german channel too... so that hits it on the nail I would say.
I'll try to find the threads you mentioned.
What comes out of this is that it helps to test with a small bit of the source to see what the result is..
I found some interlaced resizing scripts so next time I'll experiment a bit with these...

With interlaced encoding, you should get as good (or as bad) video as the material that was broadcast, provided that you've got enough bits to use. That's why it's usually best to encode as interlaced if you have a very bad source.

Have you tried to revised version by scharfis_brain?

http://forum.doom9.org/showthread.php?s=&threadid=75432

Yep for shure ... much better quality, but still some frame doubling ( very few) ... but if you got a camera move along a landscape for instance, you will notice such "jerks" (if they do appear) ;-)

So the detection of blendings has to be still dveloped so that this function will become even more acurate :)

I tried, for fun, running restore24 on one of the clips (an american music video) and the result was lots of colour bleeding and rainbows.. not very useful.

Separating the fields, processing and weaving back (keeping interlace) gave me a better result, so I think I'll go that way for the moment..