Motion adaptive deinterlacing software is more effective than you give it credit for. I invite you to try it sometime and view the results frame by frame in something like Virtualdub. Please note that I said reducing 1080i to 720p ... that has implications for the edges you are concerned about.

Deinterlacing 30P? Interesting concept....

^^^30P is always delivered via 60i with these camcorders and BD media.

Regarding 1080 vs. 720P, as I've said earlier in this thread, if you're happy with 540 lines of real rez, then you may as well go ahead and stick with 720/60P and get real HD video.

Aaron, you are totally incorrect regarding 720p vs 1080i resolution. There is unquestionably greater resolution in a 1080i broadcast. It's measured and it's observable. I'm surprised at this late date that people still think you're only getting 540 PIXELS of resolution. You also forget that resolution is measured in two directions, horizontal and vertical. Both those numbers are higher with 1080i and thus the significantly higher resolution for 1080i.

Now if you want to make a case for the resolution dropping during periods of rapid movement in a 1080i broadcast, you're now on firmer ground. However, what is not widely known, is that well over 90% of scenes on broadcast TV have little to no movement...yes, even during sports telecasts. So this is why people with 1080p TVs (like myself) can easily see the added resolution provided by 1080i.

Yes, 1080p would be the best, but it sure isn't happening anytime soon when broadcasters barely have the bandwith for 1080i.

I found another interesting web page about deinterlacing, including some information about how deinterlacing material that was originally progressive can degrade it (i.e., deinterlacing 30P material that has been converted to 30P in 60i):

http://www.100fps.com/

This is the first site I've seen that shows that there are both deinterlacing methods that result in 30 fps video (as Aaron and I have been saying), *and* deinterlacing methods that result in 60fps video. Bobbing+rescaling and bob+weave both can result in 60fps video, with bob+weave being better because it's motion adaptive and does a better job of preserving vertical resolution in static areas.

So, maybe I've been wrong - if a TV uses a motion adaptive method that includes bobbing, such as bob+weave, then it is possible that these TVs actually generate a "synthetic" 60 fps from the two fields, and thus 60i video would look more fluid on a 60p display than 30p 2:2.

There have to be some test video patterns one can generate to help figure out what a given TV is doing.

I think there is one other area in which 1080i would have reduced resolution even in static scenes. Broadcasters often blur interlaced video vertically so that sharp horizontal lines don't twitter on interlaced displays.

So when comparing 1080i to 720p, the improvement in horizontal resolution may be greater than the improvement in vertical resolution.

You could be right Dave, but the disparity in resolution between 720p & 1080i is actually much greater in the horizontal, so that again provides the argument for 1080i's picture quality advantage.

Ken, re-read what Graham is doing. He is taking each 1920x540 field and instructing his video processor to scale that field to a 1280x720 frame to obtain 60 frames/sec from 60 FIELDS/sec. I suppose that's ok from a consumer cam POV; but I still stand by my original statement that you may as well use a 720/60P cam to start with then (understand they don't really exist in the consumer space and that's too bad).

Aaron, yes, from that perspective I agree with you. I was arguing the case for 1920X1080 vs 1366X720.

Dave the 100fps site has some things wrong. IT implies that all digital video is recorded at 30fps by combining the fields in the camera before recording to tape. This is not true. The DV codec has the ability to compare fields and if there is no perceived difference/movement will combine into a frame to save data recorded to tape and allow more bandwidth for other areas. But normally the recording is field based not frame. Sort of makes me suspect the other information on the site. For more detailed info on DV etc see http://www.adamwilt.com/DV.html
I am still trying to find info on displays that say how wthey de interlace, specifically not marketing stuff. It is clear there are big differences as playback from my FX1 or SR11 to Panasonic or my iARt CRT are very different than an SD DVD from the same sources played back from a Sony DVD player or from the PS3!!! The PS3 upscales to 1080P30 and although the image has more detail it has a lot of Judder!! Haven't had the time yet to play around with different encoders etc.

Ron Evans

Ron, yes, I do agree that the site may not be entirely accurate, but I think the main thing I pulled from that site is that it is possible to do deinterlacing in such a way that you obtain a synthetic 1080/60P video stream (60 fps). I had always been assuming that all deinterlacing results in a 30P video stream which is then always displayed in 2:2 pulldown on 60Hz TV's.

I think what this brings to the argument is that although nobody here is 100% sure at the moment, it is at least possible that *some* TV's *may* be doing interlacing in such a way that they generate 60 fps frames from some combination of bobbing, scaling and weaving the fields.

If some TV's do this, it could certainly explain why 60i seems more fluid to many people, because possibly it is on their particular TV. It could also explain why you get such different results from the same source material depending on where the deinterlace is done and how it was done.

I'm also going to play around a bit with encoding. I like the idea of using 720/60P as a temporary stop-gap until most TV's have good motion compensated de-interlacing, as most TV's have good scaling even if they have poor deinterlacing. I just think you might get a better result by starting from *properly* deinterlaced 60i, which can be done on a PC (deinterlace to 1080/60P using motion adaptive bob+weave) and then down sampling to 720P rather than starting with 540 line fields and up-sampling to 720P. If I get it all working, I'll post here.

For those familiar with AviSynth, here's a simple 1080i60 to 720p60 avs script to play with:

AVIsource("02.avi")
assumetff
tdeint(mode=1)
#use mode=0 for 30P or mode=1 for 60P
lanczos4resize(1280, 720)

There are better deinterlacers, like McBob, but I find TDeint to be a good tradeoff between quality and speed. It's downloadable here: http://bengal.missouri.edu/~kes25c/TDeintv11.zip

Wow! What good timing. I just tried the same thing, but using Avidemux2. The deinterlacer I found to work the best was DGBob, which is motion compensated and can double the frame rate. Here was the filter stack for Avidemux2 that I used:

1. Deinterlace - DGBob with threshold 12 and frame rate doubling
2. mPlayer resize to 1280x720 with lanczos interpolation
3. mSharpen with strength of 70 and threshold of 15

The small amount of USM (step 3) was to add a small amount of EE as pre-compensation so that edges are preserved a little better on upsampling to 1920x1080. After upsampling, the EE is no longer visible but the edges are sharper.

I was surprised how well it worked. the DGBob deinterlacer worked well, the resulting 60 fps frame rate gives smooth playback even on a PC, and the little bit of EE caused the edges to stay fairly sharp even after upsampling.

This type of video can be distributed on BD, and might look better than 1080/60i or 1080/30PsF on lower end TV's that don't interlace well.

Dave, you can include playback devices in that category too, as well as AVR's. This is all a very complicated mess that unfortunately the consumer now has to deal with because poor decisions were made in the past. It's really too bad that you either have to become an expert on practically every CE subject today, or find an authoritative review before buying just about any piece of gear now.

The variations that you speak of are likely the result of proprietary video processing algorithms employed by the various chipset manufacturers in an attempt to differentiate their products from one another. And that's a good thing; but the flipside is it creates this vast "unknown" regarding your distributed project if you try to chart virgin territory.

I will continue to bug these manufacturers and try to get answers to these de-interlacing questions WRT their video processors and the implementation of those processors in certain key A/V components. Any info I get from them will be added to this thread...

Yes, this is why I haven't bought a TV yet. Obviously I'd like to get one that does a good job of deinterlacing both 24P 2:3 and 60i (perhaps even generating a good quality 60P stream from the 60i), but even though I'm a technical person and have some idea of what is technically possible, it's so hard to actually find out what products actually do what, and I don't have 8 hrs. a day do devote to research.

Having said that, there seems to be more standardization in how the camcorders behave, e.g., they all generate 1080/60i in more or less the same way (ignoring details like rolling vs. global shutter), and some can generate 24P in 60i and 30P in 60i. So I'm thinking the correct order is to first get a camcorder, then take the camcorder with you and view the video on the different TVs. You can't determine everything this way, such as how the TV handles true 24P or true 30P, but you can at least compare how well they deal with interlacing.

I suppose the good news is that even though interlacing is not going away any time soon, the processor chipsets are getting better and in a year or two most TV's will be able to deal with it (and 30PsF) much better.

Hopefully you'll hear something back from the chipset makers ;-)

I was told that in a choice between 60i and 24p, use 24p for more professional looking work. It's less clear whether one should use 30p or 24p, but as for the effects, 30p will produce a nice, clear, progressive image that produces less motion blur than the 24p recording would, without interlacing.

So, I'd say use 30p if you have it, and if you don't, 24p if you're shooting on a tripod and know to make slow, fluid movements, 60i if the subject's more unpredictable or you're operating handheld.