Is it harder to encode DV into MPEG2???

[edDV]

...

Suggesting that an mpeg will be inherently poorer quality that a DV file because the size has reduced from 25MB/sec to 8MB/sec seems logical. But, it isn't as simple as that. A DV file contains every frame in it's entirety, an mpeg only contains an entire frame (I frame) every so often (every 12 frames for PAL and every 15 for NTSC), the others are simply instructions to tell the decoder what has changed since the last I frame. With a static scene, this can result in a huge reduction in file size with no degradation whatsoever.

Which is precisely the reason you want to use a tripod to keep pixels identical over multiple frames for stationary backgounds. For pans, a tripod makes it easier for the MPeg encoder to detect motion of similar blocks of pixels further helping quality and compression. Handheld stationary shots and pans include both X-Y and circular motion that can confuse the encoder. This will both reduce quality and decrease compression.

A zoom puts all pixels into motion limiting compression opportunity.

[edDV]

Ulead, like Sony Vegas, Adobe Premiere/Encore and others use the Mainconcept OEM MPeg development kit and each adds their own secret formula for optimizing their encoder presets. I'm happy with the Mainconcept family of encoders for DV to MPeg2 encoding. I'm also happy with Mainconcept for MPeg2 to MPeg2 re-encoding.

Out of curiosity, do you know what those "secret" formula optimization settings are?

For DV, are you using the mainconcept defaults for GOP sizes (I believe PAL=12 and NTSC=15)? Or are you tweaking some settings in the advanced section of the encoder?

I was refering to what Sony, Adobe and ULead do to creare their preset defaults. If you get the Mainconcept developer kit (SDK) you will see that the software manufacturer has many more controls than are presented in the "Advanced" menus.

Sony Vegas seems to present the largest choice of advanced settings for the users to tweak, ULead products seem to give the user fewer advanced levers. This isn't necessarily a bad thing since most of the settings can do more harm than good.

I don't claim to be an expert on advanced encoder tweaking. I'll leave that to others.

[edDV]

Seems like a crucial issue in encoding quality has to be the nature of the source: a commercially produced movie starts with 4:4:4 video to encode (i.e. YUV sample for every pixel). MiniDV is 4:1:1 -- a luminance sample for each pixel, but color info for groups of 4 pixels on a scan line. That gets encoded to mpeg2 which (AFAIK) uses a luma sample per pixel, but color info for 2x2 groups. Perhaps the amazing thing is that MiniDV encodes as well as it does. I wonder if anyone has done a test of encoding 4:4:4 into mpeg2, then subsampling it to 4:1:1, and encoding that to see what gets lost in the process. Don't think I've ever seen it.

I do have lots of experience with 4:4:4:4 (last 4 is Alpha).

Most of the quality you see in a picture is the first 4 which is luminance. This gives you the 720x480 raster that contains almost all the perceived picture resolution.

The eye isn't as sensitive to color resolution especially in blues and reds. As a result 4:1:1 or 4:2:0 are perfectly adequate for color resolution in the final product.

4:4:4 is usually only seen in computer graphics animation and 1st generation film to video transfers.

There is high end 4:4:4:4 gear used for advanced compositing. The main advantage is the same digital filter can be used in all 4 channels and the pixels are always in the same X-Y-Z space.

Almost all studio and effects processing is done with 4:2:2 or 4:2:2:4. The higher color resolution is useful for effects processing algorithms and helps keep multigeneration spacial errors to a minimum.

Once the editing and effects are done. 4:1:1/4:2:0 is perfectly adeqaute for display and distribution.

As for MPeg2 encoding (assuming a component source), luminance and chrominance are calculated separately. The perceived resolution of the picture will come from the luminance Y. The exact registration of the U and V pixels has much less to do with perceived image quality.

It can be argued that DV is not the best aquisition format if you intend to do heavy effects processing. That is why the pros will use Digital Betacam, HDTV, or high resolution film.

There are also 4:2:2 (DVCPRO-50) and HD (DVCPROHD100) variations of the DV format used for acquisition.

[Phantom Of The Opera]

Q for Edmund Blackadder (and all the others who are recommending Procoder): do your remarks apply to Procoder Basic as well? Right now, the Basic version is the only product I can afford, and I will only consider getting it if it offers noticeably better results than TmpgEnc Plus (which is what I currently use. And I'm quite satisfied with it).

[Edmund Blackadder]

Q for Edmund Blackadder (and all the others who are recommending Procoder): do your remarks apply to Procoder Basic as well? Right now, the Basic version is the only product I can afford, and I will only consider getting it if it offers noticeably better results than TmpgEnc Plus (which is what I currently use. And I'm quite satisfied with it).

I was talking about ProCoder 1.5 and ProCoder 2.0. Version 1.5 is a little better for DV than version 2.0, as it's been discussed before. But version 2.0 is still miles better than TMPGEnc for DV to MPEG2. I've heard that the basic version is based on 2.0 engine and while giving you less tweaking options it still produces pretty much the same quality results as the full version. So you should definitely see a major improvement from TMPGEnc, even with basic ProCoder. The ProCoder encoding engine is good in any incarnation.

TMPGEnc from what I remember is very clean on steady interlaced shots (better than CCE), but once the motion is introduced the blocks start to appear. ProCoder is superior for those motion shots, with no blocks anywhere to be seen, at least with bitrates 4000kbps or higher.

[jeanl]

Hi all,
Well, I made some tests over the week end, and the more it goes, the more I realize that I might be missing a few things here.
So far, I've had great results using HC (free encoder), after de-interlacing the DV video. If I run HC on the interlaced video, I get the same kind of results as with TMPGenc, or CCE. If I deinterlace (even with a simple bob() in avisynth), the quality goes up tremendously when watched on a frame-by-frame basis on my PC. Way fewer blocks and no combing. A bit of smoothing though.
But this prompts more questions, and maybe you guys can answer them.
- Does it make sense to inspect an interlaced video on a frame-by-frame basis? I use PowerDVD and I'm wondering what it does to show you 1 frame when the material is interlaced? Is there any kind of processing to best mesh the 2 fields? Same question on a TV with a DVD player in "pause" mode... How is the image reconstructed from the 2 fields?
- Is it normal that encoders should work so much better when fed a de-interlaced video stream? I'm suspecting (but I haven't tested that) that TMPGenc, CCE or Procoder would also work much better if I de-interlaced the video prior to encoding. Does anybody have any experience with that idea?

In any case, with the de-interlacing, I'm quite happy with the results. Again, on a frame-by-frame basis, the MPEG image is nearly as good as the DV one. In normal playback mode, it's very hard to tell the difference.

jeanl

[junkmalle]

jeanl, Why don't you post a few sample frames or crops that show what you started with and your different results?

Does it make sense to inspect an interlaced video on a frame-by-frame basis?

Depends on what you're looking for. Most computer DVD players will bob deinterlace on the fly while playing. I don't know what PowerDVD does when paused. Set top DVD players normally bob deinterlace when paused (showing only one field).

Is it normal that encoders should work so much better when fed a de-interlaced video stream?

Since you're talking about going from DV to DVD MPEG I assume you reduced the frame rate back down to normal (AVISynth Bob would have doubled it). By doing so you have thrown out one field of each frame and replaced it with pixels interpolated from field that's left. You have essentially lost half the vertical resolution. So yes, what's left is easier to compress.

[jeanl]

Yes, what you say makes sense...
I'm beginning to understand my problems a bit better! I did get the frame rate back down before encoding as you mentioned.
I think my problem is that I'm not watching the output on the target display (a TV!). But I have a question:
- When I watch the original interlaced DV in PowerDVD, I don't see any combing, everything looks fine.
- When I watch an MPEG-2 encoded interlaced mpv (renamed to mpeg) in PowerDVD, I can see a lot of combing. So I'm guessing PowerDVD isn't de-interlacing that correctly.
I think I need to go back and check that first!
When I sort through my problems, I'll post some frames to show you what I'm seeing!
jeanl

[Cornucopia]

I know that my company primarily uses tripods, steadicams, etc. along w/ occasional lighting, but occasionally we do shoot handheld--on DVCam (Sony PD-150),DV (Panasonic VX100, and sometimes other cheesy consumer cams) or Betacam. This is often edited and then output to DVD. Stuff looks GREAT! on interlaced TV monitor.
Sometimes go DVD Recorder route (Pioneer PRV-1900), sometimes Canopus/Amber hardware encoder, sometime TMPGEnc.

ALWAYS looks smooth and clear w/ very little loss from MPEG2 encoding. It can be done!

You guys must be forgetting something to be getting the kind of problems/artifacts you keep mentioning.

Scott

[jeanl]

Cornucopia, I think you're right. There must be something I'm not doing right!
jeanl

[Soopafresh]

Kinda cool how well those free MPEG encoders work. HC, QuEnc, NuEnc are all top notch apps for standard DVD resolution video.

[edDV]

...
- When I watch the original interlaced DV in PowerDVD, I don't see any combing, everything looks fine.
- When I watch an MPEG-2 encoded interlaced mpv (renamed to mpeg) in PowerDVD, I can see a lot of combing. So I'm guessing PowerDVD isn't de-interlacing that correctly.

jeanl, Are you watching any if these on a burned DVD played to a TV?

As I read back through your posts, it seems your concerns are about interlace (or post deinterlace artifacts) on your computer screen.

There are two conflicting goals that need to be decided at the outset. Is the goal to view the MPeg2 optimized for progressive computer display?, or is the goal to author a DVD that will play at highest quality from a DVD player in interlace or progressive TV mode?

For goal 1, you need to obtain the best deinterlace your $ and time budget allow optimized for the desired compression. There are many threads here and in the Divx forums explaining the various strategies.

For goal 2, you need to keep the video interlaced, view what you are doing on a properly calibrated TV monitor and follow the DVD spec. You can view the interlaced video on a computer monitor using a player like PowerDVD for convenience, but you should not make quality judgements from the quick and dirty deinterlace that is used by PowerDVD (a combination bob and weave). PowerDVD is designed to be a playback viewer that will work realtime on a variety of computers and as such cannot achieve the quality of deinterlace that a number crunching adaptive algorithm can produce or that of hardware filters that exist in quality DVD players and progressive TV sets. In addition, there are many other issues of levels, gamma and color reproduction that make it important to use a TV monitor for quality judgements.

These two goals represent a fork in the road with different techniques for optimization.

As for isolating stills from an interlaced stream, what you are doing is a deinterlace for one frame display. There are quick low quality methods (like single field or field blend) and then there are sophisticated motion adaptive algorithms that look at multiple frames to calculate a still. Don't expect a product like PowerDVD to to be doing the latter. Many camcorders and DVD players use hardware deinterlacers (the best ones are using the Faroudja patents) to produce a still frame.

Reading back on this thread, I think most of your issues relate to viewing an interlace stream on a progressive computer display. I suggest you burn a test DVD. Am I missing something?

[jeanl]

I think you are quite right. I think I'm beginning to realize that! I agree with you that comparing interlaced video on a computer monitor using PowerDVD is a bit flawed from the start. I simply didn't realize that at the onset.
I believe there are video sources that PowerDVD does not even de-interlace (I'll check that tonight), for example, I'm not sure PowerDVD deinterlace an mpv video that you rename to mpeg! Obviously, it makes no sense to compare anything in that case!
I will check tonight: burn the output video to a DVD and check the quality relative to the original.
My goal is to produce a DVD that will play on a regular TV, so I should keep everything interlaced.

I'll get back to my tests and report later. Thanks to all for your help.
Jeanl

[Soopafresh]

Check out the free Dscaler app for a series of excellent DeInterlacers for Mpeg playback.

[edDV]

The best way to compare MPeg2 encoders is to first create a sequence of typical scenes from your DV material. Include stationary scenes, slow pans, zooms, fast action, indoor, bright sunlight, etc.

Then encode the same test sequence for each encoder and author them all to a test DVD. Include a colorbar pattern from your editor for proper TV monitor adjustment.

If you want a good backgrounder on how Progressive DVD players work, see this this article. 60 field per second material will also be deinterlaced with the hardware in the DVD player if you use a progressive 480p connection to a progressive TV (EDTV or HDTV) or in the progressive TV for 480i inputs.

http://www.dvdfile.com/news/special_report/production_a_z/3_2_pulldown.htm

[vhelp]

de-interlace vs. Interlace through software players ...

I agree.., that its not a good idea to make judgment calls on quality
through the avenue of an software player (ie, powerdvd) because it
is not giving you a true representation of your video (on a frame by
frame bases) and should be avoided when making test, among other things.

proper encoding and color space ...

Before futher video judging, the user should be certain that the
correct color space is the same as the source. In other words..
..after an encoding to a finalized MPEG-2 file, (assuming encoded w/
interlace) that mpeg-2 's color space should match that of the source
DV file. There should be no differeniations between the two.
.
After that (assumed to be correct) then make your further analysis
with your video judgmentings etc.
.
.. personally, I would not worry about the 4:1:1 vs. 4:2:0 differences
.. when you encode to MPEG-2, because there is no issue, since you are
.. converting from one format to another. Why is it that you don't
.. notice any differences between the 4:1:1 (dv) and 4:2:0 (mpeg-2) ???
.. ..and the simple answer is this. It's because of the color space
.. conversion algorithem from RGB_to_YUV, when conerting your DV source
.. to MPEG-2 source. When you convert your dv (yuv space) to mpeg-2 (yuv space)
.. the color space is the same, under yuv. It's the equation numbers that
.. seem to be adding up the same, between the two different formats (dv vs. mpeg)
.. and because of the limitation of the eye's ability to differiate between
.. the intensits of the YUV conversion, there is no noticable difference
.. that the eye can see.
.. .
.. As evident, you can take (asuming you encode correctly) a DV and encode
.. it to MPEG-2 file, and examine the frames from each source (dv vs. mpeg)
.. and notice no change in color space. (If you need to see a demonstration
.. of this, feel free to ask, and I'll post a sample, to help clarify)

When working with a DV source, the user should always bare in mind, that
the finalized product should be Interlaced. That means, encoded as Interlace
MPEG-2.

If you want to disect a given source (mpeg-2) you should open it inside
vdubMOD and scan through the time-line until you find a given frame that
meets your research criteria, and extract it for further analisis.
Also, when you are comparing, be sure that you are comparing the dv vs. mpeg
sources, and not the TV out of your dv device, because your dv cam (and your
tv set) is not your end source.., but rather, your dv source, that comes
out of your dv cam (via firewire) is what you want to compare with.
So, when you compare, be sure you are comparing your dv cam's dv avi sourc
file against your fianlized encoded MPEG-2 source file
.
.. You can open your DV source file inside vdubMOD (and ignore the error
.. message (regarding audio) that vdubMOD reports) and use that as your
.. main source file, and then open your finalized encoded MPEG-2 file inside
.. vdubMOD. Scan through (each vdubMOD window) to the scene to analize, and
.. go from there. Use your imagination.

Procoder vs. TMPGenc vs. CCE vs. xxx ...

Its my opinion that Procoder (v1.5) produces the best results from an interlaced
source. In my OP, TMPG comes in second.
.
However, when using Procoder, and to make certain you use the correct
color space, the proper settings to achieve the correct color space is to
make sure you select (under Add Filter) the following:
.
* 601 Correction - Shrink Color Space
.
It seems that Procoder is reversing (or assuming) and using a higher color
space filter as the default. I don't know why. And, based upon previous
posts on this 'color space' topic, this is the reason for most user's
poor results.
.
All sources, when opened inside Procoder, should have this (above) filter
included in there encoding projects. There should be no other filters
(color space wise) set inside Procoder. However, there are times, when
the user changes the color space in another utility (ie, AVIsynth) and
things get messed up (or user forgets to change back) and you have to
do some things, like not the above filter, for one. But, that is
the only time. I'm guessing that the majority of users are not using
Procoder correctly, or just don't have a better understanding of the
color space issues that go on.., and I believe its on account of the
popularity of AVIsynth scripts that get thrown around so free-willed
and without any warnings associated with them .. (because of the lack of
knowledge in the above) Anyways.

Another tip I guess ...

Is to not use filters. And understand that all software encoders out
there are RGB geared, and not YUV (as many seem to believe CCE is YUV,
I don't) Everything that is fed inside an editor (or encoder) is
always going to be RGB. That's just the way it is.
.
.. To demonstrate.., try feeding an MPEG-2 (which is YUV, not rgb) and cce
.. won't be able to read it. But, open this same MPEG-2 inside vdubMOD, and
.. frameserve it (.vdr) to TMPG (which is rgb) and save_as (.tpr) and then
.. feed this into VFAPIConv.exe and make your psuedo-avi file, and now open
.. this inside CCE - - it magically works.
.. .
.. another demonstration.., try feeding your DV (yuv) source file into CCE,
.. using the same steps as in the above, and you will meet with the same
.. success.
.
Weather your source is RGB or YUV, if it's YUV, it has to be converted to
RGB, and then it can be worked on. If you understand how MPEG-1/MPEG-2
is encoded to (from avi to mpeg) you would understand that it can't work
with YUV numbers.. they have to be RGB first, then that RGB is DCT'ed, and
then Quantized and then RUN Length Encocoded, to compress the pixels..
and then finally, the color space is converted to YUV for MPEG format.
.
The main reason why CCE is not able to open most AVI source files is
simple because it is not RGB, and because most AVI source files are
in another codec type compression. As far as I know and understand,
CCE only understand RGB format avi files. I could be wrong
but the above make more sense to me (at least) thus far

Interlace vs. progressive (as seen on your tv set) ...

From my understanding and researching, when you pause a frame from your
dvd player or vcr, and you view the image, ..when you see a fluid
and clean (non-interlaced) screen, you are not seeing a fluid and
clean screen. The screen is showing your source interlaced, but at
60 times a second, so fast, that you can't noticed it as interlace (or
weared lines) and you are fooled into thinking that it is progressive.
No matter how many times you press the pause button, or frame by frame,
scan through your video, and view on your tv set, you'll always see a
smooth clean progressive frame, but it is actaully slit and interlaced
60 times a second.
.
To test this for yourself, using a digital camera (not a dv cam) take
a picture of that same paused frame on your tv set, and share with us
all here, what you see

-vhelp 3268

[jeanl]

thanks a bunch vhelp! That's a lot of info right there! I'll try to digest all this. I like the suggestion about vdubMod to compare the original DV video and the encoded MPEG one. I haven't done that so far.
But do you know which de-interlacing technique vdubmod uses to present a single frame of an interlaced video? (just out of curiosity)

Thanks a bunch for the massive info. I'll post back after my latest tests.
jeanl

[junkmalle]

From my understanding and researching, when you pause a frame from your
dvd player or vcr, and you view the image, ..when you see a fluid and clean (non-interlaced) screen, you are not seeing a fluid and clean screen. The screen is showing your source interlaced, but at 60 times a second, so fast, that you can't noticed it as interlace

I don't know about your DVD player(s) but every player I've owned performs a drop frame deinterlace when paused. They either display the same field twice or replace the second field with data interpolated from the first.

If you want to test your player(s) burn this test image (do not deinterlace, deflicker, or resize) as a short video on an interlaced 29.97 fps NTSC movie DVD:



When you play it back on your NTSC DVD/TV it should flicker like crazy. When you hit pause you will see something like this:



The black and white halves may be reversed depending on which field your DVD player drops.

I can provide a DVD ready MPEG file if anyone wants it.

[troyvcd1]

where is this 601 colorspace adjustment in procoder 1.2

[thor300]

vhelp, "To demonstrate.., try feeding an MPEG-2 (which is YUV, not rgb) and cce .. won't be able to read it. But, open this same MPEG-2 inside vdubMOD....".

Are you mixing up YUV and YV12 or ..?

You are right that CCE will read RGB, but it also reads YUV. You can set up the registry to make it look like it also read YV12, but its actually just another DLL that is called for the colorspace conversion to YUV or RGB.

[Edmund Blackadder]

However, when using Procoder, and to make certain you use the correct
color space, the proper settings to achieve the correct color space is to
make sure you select (under Add Filter) the following:

* 601 Correction - Shrink Color Space

All sources, when opened inside Procoder, should have this (above) filter
included in there encoding projects.

Vhelp, you're correct for the most part. However, do not apply 601 Shrink filter when encoding DV AVI files (at least the ones based on MainConcept DV, Canopus DV or Microsoft DV codecs). ProCoder properly preserves DV color space when encoding to MPEG2. The 601 Shrink filter is needed however for AVI files such as Uncompressed, Huffyuv, DivX, Xvid, etc.. But do not use it for DV to MPEG2 encoding.

[jeanl]

Well, I went back and made more tests. This time, I burned a DVD with all my converted files and watched on my TV. Not surprisingly, it was very hard to tell the difference between the various encoders (the bitrate was 7.5Mb/s, pretty high). So it appears my problems all came (as suggested by an early response) from comparing my interlaced material using PowerDVD on a frame by frame basis. Whatever algorithm PowerDVD uses to de-interlace when you freeze a frame was showing very ugly stuff on some of the encoded frames, but that really does not show on my TV (in full motion of course).
This is mostly good news, as it tells me that I don't really have any problem encoding DV as I thought I did. The bad news is that it makes it kind of hard to choose encoding parameters, if what you need to do is burn a DVD and watch it on the TV. I tried to compare the various encoded files with the original using virtualdubmod (which can open the original AVI as well as encoded MPEG files). It's great for steady frames where you don't have much combing, you can really compare the various outputs. But for motion frames (the most interesting ones to compare), the combing makes it nearly impossible to compare images (it almost entirely mask the blocking artifacts that might exist).
So now I would be interested in ways to compare 2 outputs on a field by field basis. Do you guys know if there's a way to do that? I'd like to be able to compare the top field of a given frame in two encoded files, side by side. Of course, the vertical size would be half, but at least I would be able to see encoding artifacts in that case. I haven't seen that option in vdubmod... Any pointers?

In any case, thanks a whole lot for your help. Much ado about not much as it turns out, sorry about that!
jeanl

[Soopafresh]

Here's a way I discovered from going through the AviSynth Documentation

https://www.videohelp.com/forum/archive/t263749.html

It works better on shorter clips - stays in sync that way.

[jeanl]

I see, I wasn't sure that vdubmod could open avs scripts. This is a good example. I can probably find a way to modify it to show only the top or bottom field...
Thanks!
Jeanl

[junkmalle]

now I would be interested in ways to compare 2 outputs on a field by field basis. Do you guys know if there's a way to do that?

Start two instances of VirtualDub(MPEG2 or MOD). Open your two files. Add the standard Deinterlace filter:

Duplicate Field 1 or 2: will throw away one field and replace it with a copy of the other. This maintains aspect ratio.

Discard Field 1 or 2: This leaves you with one field and a half high image.

Unfold fields side by side: shows both fields side by side - a very wide, half height image.

View your results in the output pane of both instances of the program. Use Options -> Swap Input/Output Panes if necessary.

[jeanl]

now I would be interested in ways to compare 2 outputs on a field by field basis. Do you guys know if there's a way to do that?

Start two instances of VirtualDub(MPEG2 or MOD). Open your two files. Add the standard Deinterlace filter:

Duplicate Field 1 or 2: will throw away one field and replace it with a copy of the other. This maintains aspect ratio.

Discard Field 1 or 2: This leaves you with one field and a half high image.

Unfold fields side by side: shows both fields side by side - a very wide, half height image.

View your results in the output pane of both instances of the program. Use Options -> Swap Input/Output Panes if necessary.

junkmalle, that's exactly what I ended up doing and that works flawlessly. I was able to compare original and encoded field by field, exactly what I needed. Thanks for the pointers.
From my tests, I'm finding that down to 5Mb/s, HC does as good a job on my videos as CCE does. I'm really impressed. At 6Mb/s, the differences between original and encoded are quite minute, and there's just no way I could possibly see them during regular playback.

So thanks to everybody for your help! I'm a happy camper.
jeanl

[edDV]

Just one last word since you say you aren't seeing differences on the TV. Within a very few years, most people will be viewing on a EDTV or HDTV. Small encoding differences show on these sets.

Add to that the curse of the videophile. When you know what to look for, the artifacts will jump right out at you. If you intend your DVD to be useful into the future, err a bit to greater quality encoding.

[jeanl]

Just one last word since you say you aren't seeing differences on the TV. Within a very few years, most people will be viewing on a EDTV or HDTV. Small encoding differences show on these sets.

Add to that the curse of the videophile. When you know what to look for, the artifacts will jump right out at you. If you intend your DVD to be useful into the future, err a bit to greater quality encoding.

I'm 100% in agreement with you on that one! I know it won't be that long before I migrate toward a projector or a much better TV, and I want to plan ahead for that. I think I'm going to encode at 6Mb/s or more, which should give me over 1 1/2 hour on a DVD.
One question though: since ultimately given some years, these videos will be viewed on progressive displays, is that an argument for de-interlacing the video? I'm thinking I shouldn't do that since that would be throwing away info and progressive TVs or projectors can display interlaced material very well... Any thoughts?

jeanl

[edDV]

Just one last word since you say you aren't seeing differences on the TV. Within a very few years, most people will be viewing on a EDTV or HDTV. Small encoding differences show on these sets.

Add to that the curse of the videophile. When you know what to look for, the artifacts will jump right out at you. If you intend your DVD to be useful into the future, err a bit to greater quality encoding.

I'm 100% in agreement with you on that one! I know it won't be that long before I migrate toward a projector or a much better TV, and I want to plan ahead for that. I think I'm going to encode at 6Mb/s or more, which should give me over 1 1/2 hour on a DVD.
One question though: since ultimately given some years, these videos will be viewed on progressive displays, is that an argument for de-interlacing the video? I'm thinking I shouldn't do that since that would be throwing away info and progressive TVs or projectors can display interlaced material very well... Any thoughts?

jeanl

Bottom line, keep interlaced. The deinterlacing chips in the DVD players (especially the nextgen high def players) and TV sets are far superior to what you can do now in software.

Deinterlacing at full resolution is a complex subject. Most software deinterlacers are targeted to extreme compression. At full resolution they degrade the picture significantly.

Full frame high quality deinterlacing (e.g. with multiframe motion adaptive compensation) has been done in broadcast special effects equipment for 20+ years. That is the technology that is being built into the DVD player and HDTV set chipsets.

[jeanl]

Great! Thanks a bunch for the advice!
One question though: can HDTV and progressive-enabled regular TVs display interlaced videos in a "backward compatible" manner (i.e., each field every 1/60th of a second for NTSC) or do they have to de-interlace the video. I'm guessing that they are backward compatible. If they are, is it normally better to display the native interlaced stream, or do you gain something by a high-quality de-interlacing and a progressive display?

jeanl