MPEG-2 Encoding Program for DV

[midiguy]

hello. After reading some threads here, I've gathered that procorder or CCE is best to use when encoding MPEG-2 video from DV footage. but my questions is, what if you are interested in max quality that DVD can offer (1-pass CBR, max dvd spec. bitrate). What would be the best program then? (where the "best" program is the one that offers the best quality, and doesn't have to be the fastest). Keep in mind that the source is specifically DV footage. I would do my own tests but I don't have my camera just yet, so I am just wondering what you guys think. (oh, and the footage would most likely be 30 fps progressive DV footage, depending on which cam I decide to get). thanks!

[malum]

CCE would still be your best bet.If speed isn't an issue then more than one pass would give better results, although it's not as important with progressive scan.

I understand though that it has an issue producing lower field first video which all DV is, so do some reseacrh into how to solve this problem

[steve_g]

with CBR the number of passes (so long as you do at least one :D )is irrelevant, yes?

[malum]

Ah didn't spot that, yes.

But unless there is enough space to do a CBR of 9000 then a VBR multipass will give better results.

[DDogg]

CCE would still be your best bet.

Just curious if that opinion is based on side by side comparison or is opinion? I don't mean that negative, just wanted the rationale for the comment.

[midiguy]

...read below :rolleyes:

[midiguy]

it is max quality that DVD spec can offer. so it is useless to use VBR, thus useless to do more than 1 pass.. I am hearing that tmpeg would give max quality for 1-pass CBR max bitrate. I wish I had my cam so I could test for myself!! thanks for the replys though! keep em coming!

[malum]

Not based on a side by side comparision, largely based on the fact that CCE is the weapon of choice for most quality freaks and that it costs $2500 compared to TMPEG's $0 so it must be doing something for the money ;-)

I'm not sure why TMPEG would be considered better for a CBR single pass.

Having said that I think you would be doing well to spot any difference between the end results of the two methods.

[DBaT]

After trying Canopus Procoder I was quite positively suprised.
Has anyone done any comparison procoder-cce-tmpg with DV material?
I don't have CCE and don't do that often MPEG-2 conversions so I don't know how these two compare to each other.

[DDogg]

Encoders that handle progressive well like CCE may not necessarily to the same quality on interlaced source. All three encoders have their strong sides. The "conventional wisdom" that is emerging seems to be that TMPG is used by a lot of people for Progressive when available bitrate is less 1600. CCE is king of the progressive encode when good bitrate is available. ProCorder is just magic on interlaced DV/capture. Of course, "conventional wisdom" around this place changes on a weekly basis :)

[Ookami]

>Has anyone done any comparison procoder-cce-tmpg with DV material?

Several people... mb1 would be one of those who are registered here.

BTW, the x is more expensive than y ergo it's better can be beaten by too much GPL software etc. And TMPGEnc Pro is not 0$.

[chainsaw135]

so far my tests have been with cce 2.50 and procoder 1.01.35.0, i have found at a bitrate of 2500 or around with 2 passes the procoder looked better. Tho that is when i used dvd2svcd as the middle man haven't done any testing with just the encoders and don't plan on it as dvd2svcd results are good enough for my eye's.

[midiguy]

meh.. that is vbr though, and more than one pass.. what I am talking about is which one has the capacity to offer the best quality @ max bitrate, slowest settings (for quality-speed trade off), etc. I ahve been hearing that max bitrate DVD w/ the setting at "slowest" in tmpeg will yield the best results.. damn, can't wait for my new cam!

[chainsaw135]

@midguy meh, yeah so i went and did more tests:)


After testing no matter what settings i used cbr 1, 2 passes, even a 3 passes on cce, still wasn't better then procoder with any of my dv encodes.

[waldok]

From my testing on interlaced sources at CBR 5000 (don't ask why, it was just an "average bitrate" test), Procoder gives better results than CCE. But, as usual, it is purely subjective and maybe some other people would find CCE results better ?

I was (and still am) very impressed by Procoder quality on interlaced material and never used CCE on this kind of sources ever since.

I don't know where the difference comes from, but I really find quality much better. Picture looks crisper, maybe it's just a question of luminance corrections or any other pre or post processing filters, I don't know, but the result is that Procoder gives very pleasant results and is pretty easy to configure.

Just my point of view.

[DDogg]

http://forum.doom9.org/showthread.php?s=&threadid=47001

This was a similar discussion that might be of interest.

[sync]

I keep reading about how Procoder is good for interlaced sources. I found it slightly better than cce on handling motion but still not good. Both look much better if I deinterlace. I'm using SVCD bitrates.

I get the impression that people are having success with Procoder without deinterlacing but I can't figure out how. It's not like there are very many settings to tweak. How are you guys doing it?

[DDogg]

Both look much better if I deinterlace.Are you doing your judging on TV or PC?

[sync]

Originally posted by DDogg
Are you doing your judging on TV or PC?

On the TV. The interlaced version looks better on tv than pc, but still not as good as deinterlaced. My only criteria at the moment is the handling of motion, since that problem is so noticeable.

[DDogg]

The interlaced version looks better on tv than pc, but still not as good as deinterlaced. Huh? :)

[waldok]

@sync

Do you have a progressive TV or something like that ?
Otherwise, I just can't see why deinterlacing would produce better results on a TV.
Did you try to play with field order settings in procoder ? I had to find out the correct combination on this. For example, with old analog camcorder, I had to select odd field first, whereas with DV source, it was even field first. Did you try different "source field order/destination field order" to see if your motion problem comes from there ? Usually, if you have a problem on motion scenes, it is likely that you have a wrong field order at encoding time.

Hope this helps.

Waldok:cool:

[bb]

You probably know my opinion: for SVCD bitrates I'd deinterlace, too. But other opinions are that you better encode interlaced if your source is interlaced, even at SVCD bitrates. If you ask me you should create CVDs (352x576) if you prefer interlaced encoding.

Procoder's field mode is said to cause trouble on some DVD standalone players. But from a quality perspective this mode seems to be superior to the "classic" frame based encoding with alternate scan.

bb

[sync]

waldok,

It's a regular TV. I'll try changing the field order.

bb,

I created an interlaced VCD that didn't exhibit any motion problems. I guess that points to the bitrate for SVCD being too low for the resolution for interlaced material. But I recall reading that MPEG-1 handles interlacing differently than MPEG-2, so that could be a factor.

I'm using DVD2SVCD to create my files. I'm trying to figure out a way to automate the creation of a number of test files in one batch. Does anyone know of a way to do this?

[zeus163]

I prefer to de-interlace as well (which seems to put me in the minority with ProCoder). I capture from my digital cable box using my digital8 camera as a middle man through a firewire and if I don't de-interlace the fast motion scenes don't look crisp to me. I will do some more testing with deinterlace and interlace and check them out this weekend (NTSC DVD encodes). Thanks for mentioning the field options as I will try that as well. I like the results of de-interlacing except if I've added text somewhere in my encode. So, if I don't need it, then I won't use it. The joys of encoding.

[SomeJoe]

Originally posted by sync
I created an interlaced VCD that didn't exhibit any motion problems. I guess that points to the bitrate for SVCD being too low for the resolution for interlaced material. But I recall reading that MPEG-1 handles interlacing differently than MPEG-2, so that could be a factor.

sync,

It is always preferable not to deinterlace since deinterlacing removes temporal resolution.

However, since you are creating an SVCD, you may be happier with the trade-off that deinterlacing would provide.

At the very low bitrates of SVCD (2.5 Mb/sec), encoding interlaced material will be difficult for any encoder. There are 60 interlaced fields/sec to deal with, and not a lot of bitrate room for the motion vectors.

Deinterlacing reduces the effective frame rate to 30 progressive frames/sec, which is far easier for the MPEG encoder to encode. The 2.5 Mb/sec can be more efficient with it's use, and you will get less MPEG-2 artifacting such as macroblocking and mosquito noise. This is at the expense of temporal resolution.

The MPEG-1 VCDs you previously created essentialy did this, because MPEG-1 does not support interlaced video. MPEG-1 is always deinterlaced before encoding to make the best use of it's very low bitrate.

Try a good deinterlacer like some of the ones available for AVISynth, and encode with a good progressive encoder like CCE, and you will probably get the best that you can get out of 2.5 Mb/sec.

[bb]

Originally posted by SomeJoe
At the very low bitrates of SVCD (2.5 Mb/sec), encoding interlaced material will be difficult for any encoder. There are 60 interlaced fields/sec to deal with, and not a lot of bitrate room for the motion vectors.

Deinterlacing reduces the effective frame rate to 30 progressive frames/sec, which is far easier for the MPEG encoder to encode. The 2.5 Mb/sec can be more efficient with it's use, and you will get less MPEG-2 artifacting such as macroblocking and mosquito noise. This is at the expense of temporal resolution.
This is a little misleading, because the frame rate is 30 fps in both cases, and the number of pixels to encode is the same. The difference comes from the way the pixels are organized in progressive vs. interlaced frames.

But you're correct that there's a tradeoff:
- Interlaced: more complex encoding, thus more macroblocks/smoothing at the given bitrate, but smooth motion
- Progressive: less complex encoding, thus less macroblocks/smoothing at the given bitrate, but motion artefacts due to deinterlacing

bb

[sync]

Why is interlacing not an issue for people that rip DVD to SVCD? Don't DVDs use interlaced video?

[sync]

Originally posted by zeus163
I prefer to de-interlace as well (which seems to put me in the minority with ProCoder).

It seems to me that if you deinterlace your source, you might as well take advantge of the speed of CCE (assuming it's available to you).

[zeus163]

I do use CCE for divx files that I encode to DVD NTSC standards. However, ever since ProCoder was recommended to use for DV encoding, I like the results. I've written some basic avisynth scripts with the help of others on here--that I use with CCE (can't wait to get them all working with the plug-in server introduced to me by DDog), but I don't think they end up looking as good as the de-interlaced encodes that are produced with ProCoder. Plus, for me when I use those scripts there really isn't any speed advantage.

Of course, now I'm totally perplexed and wondering if my last batch of encodes will be any good to anybody else. The people I send them to think they look great, but could I be doing something better? I've actually got a DVD I made at home that has the same file done in ProCoder de-interlaced, three different avisynth scripts telecide(), fielddeinterlace() and fielddeinterlace(blend=false)--I think those are what they are. I'm at work so I'm trying to remember what they were. I guess I should also do one that is also ProCoder normal and see what the best one would be. I've yet to decide which of that batch is the best. I see something I like in each of them, but ultimately, I like the ProCoder one best.

I'm open to suggestions on this.

[SomeJoe]

Originally posted by bb
This is a little misleading, because the frame rate is 30 fps in both cases, and the number of pixels to encode is the same. The difference comes from the way the pixels are organized in progressive vs. interlaced frames.

But you're correct that there's a tradeoff:
- Interlaced: more complex encoding, thus more macroblocks/smoothing at the given bitrate, but smooth motion
- Progressive: less complex encoding, thus less macroblocks/smoothing at the given bitrate, but motion artefacts due to deinterlacing

bb

You're absolutely right, I didn't articulate myself very accurately.

The frame rate is identical, what I meant to emphasize was that with interlaced video, there are updates of motion 60 times per second (each field). Once the video is deinterlaced, the updates of motion drop to 30 times per second (only once each frame). This is also what I mean by the loss of temporal resolution.

[sync]

Originally posted by zeus163
I do use CCE for divx files that I encode to DVD NTSC standards. However, ever since ProCoder was recommended to use for DV encoding, I like the results.
My understanding is that people recommend Procoder for DV because of how it handles interlacing. And since are deinterlacing then you would be losing the prime advantage that it has.

I'm very new to this and I don't mean to pass myself off as an expert. I'm just bringing this up to see if it draws out any clarification on the subject.

[zeus163]

I'm very new to this as well. I've only recently started doing this so I'm still trying to find what works well and looks good. I'm doing an encode right now of my test file and will watch them all on the TV later tonight after the encoder is done and I re-author the DVD with the new stream.

I started using de-interlace as I didn't like the motion movements because it seemed to smooth it out (which it is), but as I read more at the expense of other areas (and I'm not sure I like that). All I want are quality encodes that look almost as if it was regular TV.

[sync]

Here's a link (from the links sticky at the top of the DV forum index) that really gets into the subject of deinterlacing:
http://www.100fps.com/

The first time I visited this site I missed this link which explains when you shouldn't deinterlace:
http://www.100fps.com/video_resolution_vs_fluidity.htm

There is a lot of material at these links so I'll summarize what it says that is relevant to this thread: "If you have interlaced material (like DV from your camcorder) and want to display it on TV sets (with a SVCD for example) instead of computer screens, then you don't need to deinterlace. If you wanted to play deinterlaced material, your TV screen would not be able to display it OR it would have to interlace it and you will lose quality."

Why is a TV unable to display deinterlaced material? How is that different than progressive material?

[sync]

Although I had run Procoder to check out the settings, I did all my testing with DVD2SVCD. After reading waldok's post about field order I took a look at this setting in DVD2SVCD. The default order is called 'automatic' and that's what I was using. I created two test files for upper order and lower order. Upper order had significant motion artifacts and lower order looked very good.

Tomorrow I'll compare it to the deinterlaced version. I'm embarassed that the solution was something so simple, but I'm glad that this seems to be solved.

[bb]

Originally posted by sync
Upper order had significant motion artifacts and lower order looked very good.
DV is always encoded bottom field first (lower order). That's one important thing to know about interlaced encoding.

bb

[zeus163]

Last night I made my wife watch 7 (or was it 8) different versions of encoding that I did (all on the same file). Three of them were avisynth scripts pulled into CCE and the other four were ProCoder versions. The worst results, (I have to add this was probably on our lamest TV too because that's where the DVD player is that plays +RW), were the ProCoder encodes especially the ones that I did not apply a de-interlace filter too. I'm not looking for SVCD bitrates as I'm doing these as NTSC DVD's. The regular lower/even field created weird effects during movements, switching it to upper/odd (even though I've read that DV is lower field first) created an even weirder effect around moving scenes, and then the non-interlaced seemed to be the best of the bunch (just for the field's section) although the color seemed to be the sharpest done with those the picture movements and affects was all wrong. When I started showing her the avisynth scripts we both thought that fielddeinterlace seemed to be the best.

The ProCoder encodes I've done that I've applied the de-interlace filter to play nicely on our nice Toshiba TV in the living room. Maybe it's what I'm doing, but if I don't apply that de-interlace filter for ProCoder the motions are weird and blurry or fuzzy and leave me with this awful feeling in my stomach.

I'd like to thank sync for summing that website up for me. I've looked at that site before. In fact, I'm still looking at it shaking my head and trying to get my best results. I've come to the conclusion that the shows I tape on late night are all interlaced though.

[sync]

Originally posted by bb
DV is always encoded bottom field first (lower order). That's one important thing to know about interlaced encoding.

bb

Yes, I think that's something I'll be remembering now.

[waldok]

Zzus wrote :
"Last night I made my wife watch 7 (or was it 8) different versions of encoding that I did (all on the same file). "

Well, I already tried this myself, but she fell asleep after the second one :D :D

Waldok:cool:

[zeus163]

I then had her watch it later again in the day when my other friend was here as I wanted another opinion!!! I'm so terrible.

[ulfschack]

Someone asked for a thread where similar comparisons have been made. Well I just can't believe that the thread that I myself played a part in hasn't been mentioned :D

This also applies to DV btw

http://forum.doom9.org/showthread.php?s=&threadid=50390

cheers

[cinemacraft]

I have done couple test between cce 2.67, mainconcept, and lsx encoders (all are Adobe premiere plugin)
source is dv footage captured by Sony PC100, used 1394 port of Pinnacle dv 500 card to capture dv footage to HD, and edited it with Adobe 6.5. Output targets were SVCD (resize by pinnacle package progam), Noise Reduction (Blur), and Deinterlace filter were applied.
CCE2.67.00.08 setting: 1700 1 pass CBR (same with Mainconcept and LSX),progressive was checked.
Result:

CCE and LSX resolution were about the same with ghost image (played on JVC progessive scan DVD on 55 Mit Intergrate projection tv).
Mainconcept was much clear than other, little ghost, but had interlace image on fast motion (really bad).

Tried again with CCE 2.67.00.08 with 3 PASS VBR, result was great. Resolution was about 80-90 percent compare with original footage. However, my pc has had problem with cce 2.67. It had memory issue with 20 (or longer)minutes footage, and file could not be create when it reached to the end.