Why go with quality over speed?

[lab-one]

I currently use the following to backup my DVD's:

DVD Shrink 3.0 Beta 5
Image Tool Classic .90 Build 5
DVD Decrypter V. 3.1.6.0

I have backedup 42 DVD's with a 100% Success rate. I started by following exactly what was stated in the DVD Shrink guide and have not see any reason at all to change.

However, on the DVD Backup Guides page, there is a statement that reads "DVD-9 -> 1 DVD±R - AKA REAL backups ;)Quality over Speed (using CCE)". This has had me toying with the idea of trying this method of backup. I just can't convince myself that there is any reason to forego my current methods.

Can someone tell me, is there a noticable difference between a DVD backed up with DVD Shrink vs. a DVD backedup via DVD Maestro or Scenarist? I cant see spending the cash for a CCE program and spending the time when I seem to get perfect backups with DVD Shrink.

Thanks.

[writersblock29]

You're stepping into some pretty dangerous territory with this question!
:)

Quality is always subjective. Always. For example, when I tell you that I don't notice much of a difference between CCE encodes and DVD Shrink, I'll be telling you what feels true to ME. Someone else out there is going to flame me for that opinion, citing CCE as the best thing to ever happen since sliced bread. At the end of the day... who cares? When you're watching your movie, you're really not going to care whether it took you three hours with DVD Shrink or three DAYS with CCE to make that copy. The only thing you're going to give a rip about is that it looks good. If you're happy with DVD Shrink and have had a killer sucess rate with it... gravy! Those on this forum that are religious CCE users will continue to use their methods, since they just aren't happy any other way. And that's great, too! No right or wrong in either approach.

Bottom line: If DVD Shrink is giving you results that you're happy with... stay happy! You can try the CCE approach and see what you think -- by all means! If you like it better... well, that's the beauty of choice. If not, then at least you arrived at your own conclusion about it. You're the only judge as to what you like best.

[KungFuCow]

The easiest answer is "It depends." If you're shrinking the whole movie down by over 10-15% then Id say yea, CCE is going to be better. The more you have to transcode, the better CCE is going to look in comparison. On episodic DVDs, which take almost the whole DVD, you can really see the difference.

[MackemX]

Originally posted by lab-one
Can someone tell me, is there a noticable difference between a DVD backed up with DVD Shrink vs. a DVD backedup via DVD Maestro or Scenarist? I cant see spending the cash for a CCE program and spending the time when I seem to get perfect backups with DVD Shrink.

Thanks. I can see a noticable difference :D

it's the price of DVD Maestro and Scenarist :eek:, (which one can you not buy any more?)

quality is always in the eye of the beholder and having different setups doesn't help either ;)

[Kedirekin]

I've attempted a number of transfers both ways. I found, when I kept everything the same*, I preferred the Shrink output.

Shrink introduces more macroblock noise when you push the compression too high, but typically has much less mosquito noise, and I find mosquito noise much more objectionable.

In those times when I couldn't get acceptable results from Shrink, I also couldn't get acceptable results from re-encoding, so I end up splitting to two disks or discarding some extras. I rarely even attempt re-encoding anymore.

* in other words, do the extras at a reasonable compression ratio, not 1850 kbps which is *really* nasty.

[DDogg]

Just some rambling for your consideration.

CCE becomes extremely valuable when using OPV mode (one pass VBR). OPV is based upon a constant quantizer method of encoding. Using somewhat advanced methods it becomes possible to predict the final size and quality in advance of doing the encode. Saying that in another way, it also allows seeing if a particular source will fit on a DVD at a quality level acceptable to you in advance.

Of course this method requires a fair amount of video nerdism, time, and experience to accomplish.

I would disagree with one comment made above about CCE taking a long time to encode, or words to that effect. On my middle of the road XP2800-3000 equivalent it takes about 45 minutes to do the encode using a two hour source if I don't use a lot of filters. The current methods of OPV size prediction will normally get you within 3-4% of prediction (mostly undersized). DVD users benefit more than SVCD'ers time wise using OPV (IMO) because SVCD normally needs a second sizing pass to get it just right as every little bit is needed for SVCD. A 3% error for DVD'ers is not much to worry about given the capacity of DVD so they can get the real advantage of OPV speed (which can easily reach 3.x realtime on recent inexpensive CPUs).

Speaking of filters, which I do understand is a religious topic for DVD'ers, my personal position is that the judicious use of *light* filtering can be beneficial. (I know, I know, can of worms, so don't beat me up :) ). The mpeg2dec3/avisynth/CCE method allows a lot of flexibility from a filter chain standpoint that you just can't get with a 1 click solution.

In closing I would just mirror most of the remarks above. If you are happy with the quality of your present method than stay with it. It has got to be the easiest, and least expensive, method available. If it ain't broke...:)

Good luck.

DD

[echooff]

The old cliche "to each his own" certainly comes into play. I'm a ReJig enthusiast. I'm quite happy with the output I get from this program. But then again as Kedirekin stated no one solution works "best" for me.

[lab-one]

Thanks for all of the responses. I am going to try the Scenarist NT & the Big 3 Guide and see what happens.

[ulfschack]

I've been out of the absolute "current events" for a while, so yesterday I stumbled upon a 30-pages long thread about ReJig. I read the first few, and then the last few (since my employer pays me to do something quite different). With all the praise it recevied there had to be something about this little app, I thought. So i ran it against shrink (deep analysis) and compared about twenty-something stills from a movie 7.5 gig big.

Shrink won, no question about it. Reason for it being so simple was that about four out of five frames looked identical whilst the fith had massively more macroblocking.

Now I'm not saying that every fith frame is crap from ReJig. I choose really extreem action packed caps to compare, like jumping shaking fotage from a white-river adventure (indiana jones - temple of doom).

Seeing as the video quality is either (almost) identical or noticably worse, I'm convinced that there's hope of cracking that odd crap-frame too, so i truly believe that ReJig has great potential. It's just not great yet.

At least I'm happy to know that i still use one of the best simple one-click transoders. Others might be pissed ... like always when someone draws conclusions from a comparison. Either that or the equally annoyning "to each his own"-statement, which I always find a wee bit peremtory :)

cheers

[mdkusr]

Originally posted by ulfschack
...Shrink won, no question about it. Reason for it being so simple was that about four out of five frames looked identical whilst the fith had massively more macroblocking.

Now I'm not saying that every fith frame is crap from ReJig. I choose really extreem action packed caps to compare, like jumping shaking fotage from a white-river adventure (indiana jones - temple of doom)...

Try running it through a frameserver or something. Filter out the fifth frame with it which is actually probably interlacing. Otherwise, for a direct compare, use a current gen movie (something recently released typically is pure progressive lately) with pure Progressive Frame output. This will have pure frames throughout and you won't need to run it through a filter.

Older movies tend to have the 4th and 5th frame as interlaced while the first 3 are progressive. If you apply a filter to it, you can fix it. Rejig seems to be similar to CCE in it's capability from what I read here and will require the frameserver for these frames to be filtered correctly. Once you learn how to frame serve correctly against DVD-Shrink, re-run your test.

-M

PS. Re-post afterwards since I am curious as to the real quality of this new little toy. If you have run it through a frameserver for the test posted above, I guess we will have to wait to see what happens then.

[ulfschack]

Well, yes it's run through a frame server i that respect that I create a .d2v-file which is imported into a script and decoded by mpeg2dec3.dll, then dragged into VirtualDub.

I've been doing a lot of interlaced film not to mention a helluva lot interlaced DV material, so I know what to look for. These blocks are for sure not a product of such.

Other things that support this is that the method is applied for both shrink and rejig. Also by browsing through as many pictures as I did I didn't find any resembling interlaced. TOD is an old movie, but the digital mastering of it is very recent (at least for PAL). I'm almost certain that it's 100% progressive.

No, I'm pretty sure that the blocksd are introduced aws early as in the ReJig process, but like I said it should be very poissible indeed to get rid of those crap-frames given the quality of the vast majority of the frames.

cheers

[dragongodz]

ulfschack - if you want to do a realistic comparison of rejig against shrink then use dvdshrink 1 and not the latest. why ? gee lets see, shrink has things like deep analysis, shrink has been evolving and getting better for quite some time now, shrink has a more inteligent sizing system etc etc
so whats rejig got ? its open source so hopefully more people will eventually help make it better(not holding breath though). its only new and shows reasonable results so should only get better.

also any comments on rejig would probably be better in the rejig thread so that more people involved in it are likly to read it. :)

[ulfschack]

So are you saying that ReJig is better because it doesnt enjoy the abovementioned benefits of shrink? Should I then choose to compress my movies otherwise from now on?

Can't a simple report just be absorbed whithout people flaming? Why would I wanna compare it to shrink1 ? I don't use it. I will however start using ReJig as soon as it outperforms shrink. I'm sure there are at least one dweller here interested in knowing when that time should arise. It doesn't get much simpler than that, now does it?

cheers

[MackemX]

lol :), I see both points being made ;) but I do find it strange why you would compare it against an older version you will never use again but then again DVDShrink has been around longer than rejig, so rejig will be better in months to come :)

but who's to say DVDShrink won't improve also in it's next version ;)

the good thing for all concerned is that whatever happens it will result in better quality no matter what tool you use :D

As for comparisions now and in the future, well I must use the old favourite that seems to get on some people's nerves but people can get very one sided in this area of judgement and tempers flare ;). Yes we can 'debate' about it but it won't change someone else's mind. If people are really that bothered about quality they will test it for themselves and be able to make up their own mind all by themselves

'to each his own'

as for quality over speed, I'd pick quality all the time and not just with DVD Backup but everything in general. I mean how many backups do people need to do in a day?!?!??!

p.s. apologies ulfshack

[dragongodz]

ulfschack - whats your problem ? where the hell did i flame you ?

nowhere did i say rejig was better because it didnt have those features. try reading again mate. i said a comparison against the latest dvdshrink is hardly realistic since rejig is so new and has not had time to evolve and improve the way shrink has. of course shrink is going to look and do better. if you however look back to dvdshrink 1 you will see it had quality issues to start with aswell and has improved amazingly since, hopefully rejig will do the same. is that a hard concept for you to grasp without getting defensive ?

[MackemX]

dragongodz, I think it may have been the 'Why?, gee lets see' bit :) as it could be taken the wrong way

it's a simple misunderstanding guys that's why I said I see both sides of the coin ;)

[dragongodz]

MackemX - hmm possibly. i was trying to point out obvious things that should be considered in such a comparison though. my mistake obviously.

[DDogg]

ulfshack, I do understand the point you are making, but I think you need to take a few minutes and think about the bigger picture.

Rejig is work *in progress* and is *very* important to all of us here on this forum. It is work done by our 'own', for our 'own'. The current version is a just an iteration of that contributive work by so many of the great folks of 'our' forum and our extended video community.

Nobody knows where it will end up. It may be the next killer video app or it may end up as a learning exercise. One thing is for sure, it has increased, and will continue to increase, our knowledge-base. It follows that the contributors need, and will continue to need, all of our encouragement for their efforts. Maybe we should all keep that in mind as the project goes forward.

Any single dose of constructive criticism should be accompanied by at least three doses of positive encouragement :)

Oh, as a preemptive to all, let's not let this thread turn into any type of flame.

[ulfschack]

@makemX. *lol* apology duely accepted because I believe in my heart that you also do ;)

@Dragongodz I guess it was that "gee"-part that got me, like MakemX pointed out (glad I'm not imagining things). I guess that you could see my "report" as free advice to when to start using ReJig for your permanent back-ups. Today it was "no, not yet guys". Next month it might still be "nope, but getting there so hang on" ... etc. This was my level of ambition ...yes or no. I leave the more qualified and detailed speculating to those "more involved". I still dared to forsee that the quality most likely would improve. I wonder how many angered post I would've prompted had I left that out :)

@DDogg Hey Ddogg, have you become a moderator now? Congratz :). None more deserving than you, in my eyes! Yes a ratio of three to one seems reasonable. Sorry for the 1:1 :) BTW I don't think that the fact that ReJig is very important to this community has escaped anybody's attention, I mean, will you look at the size of that thread! I don't think I've seen anything quite likely.

Oh, I just thought of a positive statement! I intend to use ReJig regardless for shrinking any 40 hour DV-encode from TmpegEnc that comes out a few megs to large. I guess I could author and then use shrink, but I'm not gonna ... at least not now ;)

cheers

[DDogg]

Kinda back on the original subject, I really like the interfaces of the one click solutions. I wish there was a way to integrate an optional re-encode mode that could be used when needed (maybe as a conditional process). That would be powerful indeed, especially if it allowed a filter chain. I think Nic was toying with that idea using a GPL mpeg2 encoder in rejig as an option, but I was never clear on whether it is something under serious consideration and whether it would deliver a high enough quality to make it worth the effort. [edited]

[Pfc Joker]

"Of course this method requires a fair amount of video nerdism, time, and experience to accomplish."

This says it all to me, one click backup solutions are easy, fast, and you can get the hang of it very quickly. Not to mention the quality is excelent for the time you spend, period.

[Kedirekin]

I hear you DDogg. I think the holy grail is a high quality encoder (TmpgEnc quality) that encodes straight from the original mpeg stream, maintaining all the GOP structure and flags (or at least has the option to). To me that is the big advantage of transcoders - you never have to worry about wrong field order or IVTC errors.

Technically I suspose this magical encoder should be called a transcoder, but it would be significantly different than any of the Compressed Domain transcoders we're familiar with.

[ulfschack]

How's that Kedirekin? I thought that transcoders in general maintains all of that structure since they have no "intelligence" in that respect, whereas encoders completely rebuilds the video. Certainly TmpgEnc has the ability to import the I-frame intevals, but transcoders do so by default and don't tamper with the stream other than shaving down some of the DCT coefficiants.

@DDogg
I don't think filter chains are at all possible w/out decompressing to "full blown" pictures if you dont want to rewrite transforms to these filters into working from the frequenzy domain, which would be quite a mind job (especially the temporal ones).

On the other hand if we do decompress (it could be conditional determined by PSNR comparison f ex) maybe this built-in encoder didn't have to be so great but could instead be worthwhile doing by just upping the bitrate temporarely (alltough not higher than the original's, for obvious reasons) to achieve better quality (with the given motion vectors at hand) than the simple transcode. Though only very limited and hard-to-code picture sequences.

cheers

[Kedirekin]

Transcoders do maintain the GOP structure, but they also maintain the structure of all the flags (especially in my opinion the RFF/TFF flags). That is their big advantage (in addition to speed). Right now, if I'm not mistaken, the only thing they do is change the quantization.

I think the next step is to have a transcoder that maintains the GOP structure and all the flags, but fully decompresses the stream and re-encodes it to a lower bitrate (including motion estimation, possibly using the existing motion vectors as a hint).

TMpgEnc with I-frame intervals isn't quite the same thing. Think of mixed footage (a stream where some of it is 23.976 fps with pulldown flags and some of it is 29.97 fps with no pulldown). With TMpgEnc, we either have to convert the stream to all 23.976 fps or all 29.97 fps - it's not possible to maintain I-frame intervals under those circumstances, and we have to make sacrifices either way. If we convert to 29.97 fps, we waste precious bitrate. If we convert to 23.976 fps (using IVTC for example) we mess up the interlaced portion of the stream (i.e. the studio intro is jumpy).

You might think that what I'm talking about isn't that much different than the Compressed Domain transcoders we're already using, but I think it is. If nothing else, it opens up whole vistas of possibilities. I can imagine re-encoding extras to a lower resolution (something I don't think transcoders will do well any time soon). I can imagine a smart IVTC that can turn a 6% NTSC stream into pure film, but without touching (and possibly messing up) the part that is already FILM. I can even imagine a smarter IVTC that can do partial inverse telecine on a 100% NTSC stream, being able to recognize where the footage is really progressive and where it is not.

Oh, and such a transcoder would also open up the possibility to introduce filters into the transcode. I've backed up a couple of DVDs that I think would have benefited from a little noise filtering and smart smoothing.

[ulfschack]

Hadn't thought of NTSC. Thank goodness I'm PAL. Otherwise I agree fully and think you have some good ideas. :)

[r6d2]

Originally posted by Kedirekin
If we convert to 29.97 fps, we waste precious bitrate. If we convert to 23.976 fps (using IVTC for example) we mess up the interlaced portion of the stream.
It's a common belief that using 30 fps instead of 24 fps leads to higher bitrate consumption. However, I have found that not to be the case.

What you lose when going from 24 to 30 is a relevant number of I-frames, meaning the error propagation inherent to MPEG is higher and hence quality (measured as quantization) is lower. But BR itself does not change. The additional B frames you create by enlarging the GOP don't really count as much in terms of BR consumption.

In fact, when it's about purely interlaced sources, IMHO they are better encoded interlaced (resized by frames and all that stuff).
I can imagine a smart IVTC that can turn a 6% NTSC stream into pure film, but without touching (and possibly messing up) the part that is already FILM.
len0x and bilu have done extensive research on the subject. This (http://forum.doom9.org/showthread.php?s=&threadid=65985) thread may be of interest.

[Doom9]

It's a common belief that using 30 fps instead of 24 fps leads to higher bitrate consumption. However, I have found that not to be the case.I don't think Kenderikin meant to imply that. He meant to say we waste bits on frames we don't really need. Obviously unless you use a fixed quality/quantizer mode, the bitrate will be independent of the framerate, but the higher the framerate, the lower the number of bits per frame and thus the lower the overall quality.

[r6d2]

Originally posted by Doom9
I don't think Kenderikin meant to imply that.
I don't think that either. I just pointed out the extent of a common belief.
but the higher the framerate, the lower the number of bits per frame and thus the lower the overall quality.
Did I say other thing?:
Originally posted by r6d2
the error propagation inherent to MPEG is higher and hence quality (measured as quantization) is lower.

[Kedirekin]

I've given this bitrate/framerate question a bit of thought. It's a bit of a mind twister, but in these particular circumstances I think r6d2 is right.

Assume a 29.97 fps GOP ½ second in length, which is therefore 15 pictures. We're keeping the structure of the stream the same, so if we IVTC the GOP down to 23.976 fps, it'll still be ½ second in length, and therefore now only 12 pictures. Since both GOPs are ½ second in length, they are allocated the same number of bits.

The original GOP has more pictures, which means fewer bits per picture, but the 'extra' pictures are one P frame and two B frames, so the wastage of bitrate is not that significant.

Interesting.

[ulfschack]

Instead of speculating you should do an objective (mathematical)comparison of these two framerates (vs original). Avisynth 2.5x has a nice built-in compare funktion that does just that. (An average over an integer number of gops would be recommended.)

cheers

[dragongodz]

DDog - Nic made a command line version of ffmpeg available to test and get opinions on the quality of encoding. the lack of response appears to have put that on the backburner for now. though down the track Nic may be encouraged to look back in to it i am sure.

both keeping original GOP structure and losing it both have advantages. keeping means you keep the way it was originally encoded so theoretically keep it more like the original. disregarding original means you can use things like scene change detection to(again theoretically) minimise any loss(especially to high motion scenes etc).