An objective 1-Click-DVD-backup-solutions comparison

[colemar]

I made a video quality comparison among these programs:


DVDShrinkHQ = DVD Shrink 3.2 deep analysis AEC sharp
DVDShrink = DVD Shrink 3.2
DVDRebuilderMO = (movie only) +
DVD Rebuilder Pro 0.78 +
CCE SP 2.67.0.10
InstantCopy = Instant Copy 8.0.0.267 aka 8.0.4
IntervideoDVDCopy = Intervideo DVD Copy 3 version 3.0B016.43C00


I know that DVD Rebuilder (that's the program I liked best before this test) is not a transcoder but nonetheless I was interested in comparing the quality of recoding versus transcoding.

The comparison was based on SSIM (Structural Similarity Index Metric) as found in http://perso.wanadoo.fr/reservoir/dl/SSIM-0.24.rar .
SSIM is an Avisynth plugin; its goal is to compare a clip with the corresponding original version; the comparison is done frame by frame and for each frame a quality index is given.
SSIM quality index ranges from 0 to 1, with 1 representing perfect match with the original.
SSIM looks far superior to other quality metrics (such as PSNR) as far as human perception is concerned.
Search for "SSIM" in doom9's forum to learn more.

Steps I followed:
- rip a movie from DVD ("The Terminal", http://www.imdb.com/title/tt0362227/, 2h:3m:31s, italian release, volume label TERMINAL_G5, ripped with DVD Decrypter 3.5.2.0 resulting in ISO file size 7,180,681,216 bytes); this is the "original" VOB file set
- make a movie-only dvd backup with each program (to remove menus and extras before running DVD Rebuilder, pre-process the VOB set with DVD Shrink set to "no compression"); each produced VOB set is now a "copy"; the original VOB set had 4972 MB of video data (the rest being audio tracks and subtitles left untouched) and each copy had about 3660 Mb of video data, that is 73.6% compression ratio (or compression 26.4% = 100 - 73.6)
- index original and copies with DGIndex 1.2.1; this is required to later open the VOB sets with the avisynth plugin DGDecode 1.2.1
- make an Avisynth script to: open each copy, syncronize with the original, select 2 frames every 500 (totalling 698 frames), feed the resulting clips to SSIM plugin

The result was a one-column csv file for each copy listing the quality index associated to each frame.

Then I imported these data files in Excel and made some statistics for the SSIM index computed over the 698 frames:


average std.deviation rank
DVDShrinkHQ 0.993520011 0.0032672 1
DVDRebuilderMO 0.991281236 0.0033843 2
DVDShrink 0.991154219 0.0046858 3
InstantCopy 0.986904668 0.0033010 4
IntervideoDVDCopy 0.985855249 0.0081932 5


The greater the averaged index, the better the average perceived quality over the entire movie copy.
When averaged index is almost the same, we should favor the program yielding less index deviation because quality is more even.

It is to be noted that no program has an index better than another program for all frames (see the attached graph: SSIM vs Frame); that is, pick any two programs and there are frames where the first program performs better and other frames where it performs worse.
This fact should sound an alert bell for those who like to pick only a few still frames to compare two programs.

I looked at specific frames (going back and forth over copy1-original-copy2 with the help of another avisynth script) and the SSIM index looks correct in respect to the perceived quality.

An explanation for the attached image:

The first graph shows SSIM index for each program versus frame number; the frames were reordered to have the SSIM index for DVDRebuilderMO in ascending order.
I believe that a graph with frames in natural order does not make much sense, since the chosen frames are scattered along the entire movie and are not correlated with their nearest frames. A more sensible strategy is to assume the SSIM index for DVDRebuilderMO as a measure for the simpleness of the frame (remember that CCE is an encoder) and thus order frames by ascending simpleness (descending complexity)!
This obviously results in a smooth ascending graph for DVDRebuilderMO, and its yellow curve acts as a sort of reference for the other curves.
I added a polinomial approximation for each curve (except for DVDRebuilderMO) to better show the trends.

The second graph shows frame frequency versus SSIM index; that is, the Y axis is how many frames got an index value that is "around" the index value listed on the X axis.
This yields an acceptable approximation of the theoretical probability density function.
The graph can suggest more information than what can be inferred from the average SSIM index (position of the curve barycentre) and std.deviation (spreading of the curve).

http://img2.uploadimages.net/324676SSIM-Graphs.png

------------------------

17th March 2005 at 19:34

For those who care, here is another test which compares only DVD Rebuilder, DVD Shrink and Nero Recode 2.
I hope this test is more meaningful than the previous.


DVDShrink = DVD Shrink 3.2, deep analysis, AEC sharp
DVDRebuilderMO = DVD Rebuilder Pro 0.78
with default settings for CCE:
VBR_Bias: 25, VBR_Passes: 2, Quality Prec: 16
+ CCE SP 2.67.0.10
NR2 = Nero Recode 2.2.6.9, Adv. analysis, High Q mode


The original was the same VOB set as before but this time I added another title (a piece from the first title) to have enough data to force a stronger compression.
The original had 6902 MB of data, 5953 MB of which was video.
DVDShrink copy had 4464 MB of data, 3491 MB of which was video.
DVD-RB copy had 4422 MB of data, 3477 MB of which was video.
NR2 copy had 4458 MB of data, 3486 MB of which was video.

That is:

Video Compression Ratio
DVDShrink 58.6%
DVD-RB 58.4%
NR2 58.6%


Then I selected 60 frames every 1500, totalling 7740 frames.
60 frames should equate to 4 consecutive GOPs.

The attached graph shows frame frequency versus SSIM index; that is, the Y axis is how many frames got an index value that is between the previous and the next X tick.
This curve is more properly named as "histogram".

For example: at X=0.9945 we see that:
DVD-RB generated about 150 frames that got an index between 0.994 and 0.995
DVD Shrink generated about 200 frames in the same index range
NR2 generated about 270 frames in the same index range

The yellow, blue and red vertical lines are markers for the average SSIM index.

http://img165.exs.cx/img165/8818/freqvsssimbigdvd1fr.png

The original included some black frames (credits) that got SSIM=1 for all the programs.

[MackemX]

wow :eek:, some good work there colemar ;)

[TheSeeker]

However the results from tests using SSIM as a measurement tool must be taken with a grain of salt. The reason being that transcoders are far less invasive in their changes to the actual video stream. I mean all they are doing is changing some quantization numbers. Encoders however have to totally recalculate motion vectors and totally rebuild the stream from stractch so to speak. So of course a tool that measures how close to the original a shrunk video stream is would favor a transcoder over an encoder just because it makes much less drastic changes to the video stream. My point being that just because its farther from the original in structure similarity, that doesnt mean that it doesnt look better than the video stream that is closer (in SSIM) to the original.

[colemar]

Originally posted by TheSeeker
... My point being that just because its farther from the original in structure similarity, that doesnt mean that it doesnt look better than the video stream that is closer (in SSIM) to the original.

I agree with you on that, but "looking better" will never be measurable (except of course by means of some human like computerized vision system).

What is measurable - once you define a "distance" - is how far a frame is from the corresponding original frame.
The distance may be (1-S) where S is the SSIM index value.

And I already checked with my eyes that for about 15 frames chosen at random, the frame that has the best SSIM index is the frame that "looks better".

[TheSeeker]

Originally posted by colemar
I agree with you on that, but "looking better" will never be measurable (except of course by means of some human like computerized vision system).

This is true. Quality comparisons will be a hot topic til the end of time i think. Because what looks better is different from person to person. I love dvd shrink as a simple, fast solution to backing up dvd's. I just happen to prefer Rebuilder. Personal preference I guess. I just love the options I have (with avisynth scripts and all).

[Didée]

Although SSIM ranking is more "natural" than PSNR, its result still have not very much to do with perveived quality:

Both methods make single frame evaluation only. There is no temporal exploration.

But temporal "consistency" is the perhaps most important factor ...

Example. We compare two different recoded streams. Stream A has an average/overall/whatever error of 5%. Stream B has an error of 3%.

So, stream B is "very much better" ;) - case closed, hooray.

But now imagine this:

Stream A got its 5% through, say, a constant color shift towards blue.

Stream B got its 3% by producing quickly varying color shifts - towards red, towards green, red-green-red-green ...

Stream B got the better metrics. But anyone who is watching would prefer Stream A for sure. Stream B might have produced an error of smaller amplitude, but since its error is varying, it is much more annoying than the bigger, but constant error of Stream A.


Over-simplistic, but you get the idea. Without evaluation of temporal correlations, usage of quality metrics to compare different en-/re-coding engines is amusement.

[colemar]

Originally posted by Didée
... Without evaluation of temporal correlations, usage of quality metrics to compare different en-/re-coding engines is amusement.

Still, I believe this benchmark is an improvement over all of the benchmarks I found until today.
And I was amused in doing that too :D

[Lefungus]

Nice test. Especially the density plots.
You should have used SSIM 0.24 though, which has been out for a loooong time now.

[colemar]

Originally posted by Lefungus
You should have used SSIM 0.24 though, which has been out for a loooong time now.

Actually I used a SSIM.dll that has file version 0.23.0.0.
It even writes "SSIM: Structural Similarity Index Metric 0.23" in the summary file.

But I realize now it is byte-for-byte equal to that contained in:
http://perso.wanadoo.fr/reservoir/dl/SSIM-0.24.rar
So it must be version 0.24.

[Surf]

Interesting even though I don't understand the techno details. I hope you can include Nero's Recode in your test. I have been assuming that it has an improvement over the dvdshrink....

[jdobbs]

Sigh.... we've discussed the limits and inaccuracies of these tests ad infinitum in about 100 previous similar posts. They are always flawed, always inaccurate and violate what is intuitively obvious visually to the casual observer.

It seems like every couple of months someone new discovers SSIM and its a revelation.... but if they truly understood it, they would know why it shouldn't be used to compare across encoders...

IMHO if anyone posts a test that shows a transcoder making quality superior to an encoder as good as CCE -- it has failed the common sense test before it leaves the gate.

Of course I did wake up on the wrong side of the bed this morning.

[Sir Didymus]

Hi colemar, :)

Me too sharing all of the theoric doubts [and the common sense infringements...] on applying the SSIM metric for a comparison among encoders and decoders. Just want to add also that all transcoders keep intact the GOP structure of the source, while encoders have, in general, a different one. So you are comparing I frames to B or P frames, making not homogeneous the comparison.

Also it seems to me your work lead to some conclusions on the basis of just some few hundred frames, sampled in a very sparse way from a title containing very few high motion scenes...

Nice test indeed...

SD

[Surf]

My guardian angel's watching over me, or is it the red-suited dude...

The forum's quiet this week especially in IFO/VOB editors room. Must be my presence....:D

So, I've seen glimpses of references to dvd Rebuilder and thought of giving it a shot. Talk about complexity! Things have changed from the dvdshrink re-author days with the start-end chop chop method. Now with great gurus from the IFO/VOB editor room, I have been achieving much more satisfactory results.

My current backup strategy takes a minimum of 2 hours or more from start to finish. Decrypter>VobBlanker>PgcEdit(+Jump2Pgc)>MenuShrink >Nero's Recode>Imgtool+Decryper. Pheew, a mouthful.

So, how long am I looking for with DVD Rebuilder? I know it's kinda broad, guestimate's good.

TIA

[jdobbs]

On my Athlon XP 3200+ it takes about 2 hours for a CCE encode (2 pass VBR) of a typical DVD (2 hour movie, reasonable extras and menus). If you want to do more passes (I usually don't) add about 45-50 minutes per pass.

If you do preprocessing with other packages etc. you'd have to add whatever time that takes.

[jwo62]

1-Click-DVD-backup-solutions comparison? My preffered program is none of the above. if I have to choose from the ones listed,it would be dvshrink ,but thats not based on any picture quality standards.

while there is some difference from one program to another,the difference is not great enough to judge whether a program is "the best"based on video quality. I watch my movies on a sony wega,it is so good i can actually see flaws/pixels on the original dvds. I use dvd95copy(primarily),dvd2one,dvdshrink,IC and dvd rb. it is almost impossible to tell the difference in video quality between them. and if you have to freeze frame the picture on your monitor to count pixels,you really are wasting your time.

its like saying the fastest car in the world must be therefore be the best car in the world,completly ignoring all other factors,like comfort,handling,gas mileage and cost.

there is a lot more to a program than the quality of the video. reliability,compatability,program support,time to process and ease of use are more important than video quality,which can only really be judged by your own eyes.

The most important thing,is to simply enjoy the movie.

really, they all do a damn good job.

[dragongodz]

IMHO if anyone posts a test that shows a transcoder making quality superior to an encoder as good as CCE -- it has failed the common sense test before it leaves the gate.
sorry jdobbs but i do have to disagree with that. an easy example, if you are reducing to a target of say 95% of original size, in this case most transcoders will only touch B frames a bit. so a great percentage of the output footage will infact be the original frames. so in this case the transcoded result may easily look as good or better than a fully re-encoded result. when the percentage you are compressing down to gets much higher then it becomes a different ball game.

but "looking better" will never be measurable
thats what a blind poll is for. its the only way you would stand a chance of getting any sort of meaningful results. also "looking better" is the only true important measure, all else is occasionally amusing but will never be a final result.

[EDIT]
and yes this has been done to death before about the merits of SSIM and PSNR and comparison of transcoders etc etc etc.
end result is use what you are happy.
i have not voted because this is not a blind test with footage to download and compare, so there is NOTHING to vote for.

[jdobbs]

@dragongodz

sorry jdobbs but i do have to disagree with that. an easy example, if you are reducing to a target of say 95% of original size, in this case most transcoders will only touch B frames a bit. so a great percentage of the output footage will infact be the original frames. so in this case the transcoded result may easily look as good or better than a fully re-encoded result. when the percentage you are compressing down to gets much higher then it becomes a different ball game. Absolutely. I've said elsewhere that in circumstances where the needed compression is small, a transcoder can actually do exceptionally well. Let's be honest though. How many films have you run across that only need to be compressed to 95%?

Of course in this test "movie-only" mode was used, which I've said several hundred times isn't what DVD-RB is designed to do. That prejudices the results by skewing it more toward methods that do ok when they don't have to work hard. DVD-RB is meant to backup whole DVDs. When you do that you have to increase the compression levels -- and transcoders fall flat on their faces.

But the example you use also helps to emphasize my point. That selection of modification of B-FRAMEs only can be the source of the annoying background pulsing that is so well known in transcoders. BUT -- an SSIM test would give it an incredibly high score (since SSIM does frame-to-frame comparisons) The I and P frames would be perfect because they hadn't changed. In the real world, though, the motion video isn't very pleasing to the eye.

I swore I wouldn't get involved in another of these ridiculous test reports, but here I am...

[dragongodz]

Let's be honest though. How many films have you run across that only need to be compressed to 95%?
a few. some of the cheap dvds sold here have bugger all extras(maybe a trailer for the movie) and after you remove any extra languages etc then they some are pretty close to being DVD5's, infact some are. :)
you raise a good point though and one that is ALWAYS a major mistake of these "tests", that 1 dvds results actually prove anything. the fact that there are thoughsands of dvds with a miriad of combinations of footage types at different bitrates etc etc etc means such ridiculously small samples ,such as 1 dvd, proves nothing except for that dvd and when its SSIM or PSNR results instead of something like a blind test then it actually proves nothing.

But the example you use also helps to emphasize my point. That selection of modification of B-FRAMEs only can be the source of the annoying background pulsing that is so well known in transcoders. BUT -- an SSIM test would give it an incredibly high score (since SSIM does frame-to-frame comparisons) The I and P frames would be perfect because they hadn't changed. In the real world, though, the motion video isn't very pleasing to the eye
sometimes yes and sometimes no, depends a lot on the original footage. the pulsing is not actually a common thing, especially for small amounts. most of the complaints about that was with higher amounts of compression with dvdshrink before it had the AEC settings. now you can choose sharper settings which should not pulse but may show more macroblocks or smoother which may blur slightly and may pulse.
yes the invalidating of the SSIM results because of original frames was also another point i meant to make. ;)

I swore I wouldn't get involved in another of these ridiculous test reports, but here I am...
i know, join the club. its just so frustrating to see these done as if they do prove something so how can a person not want to set the record straight..... AGAIN ? must resist next time, must resist. ;)

[jdobbs]

It's just that people who are new to video come to DOOM9 because they know there's a lot of knowledge here. Then they see something like this that looks so damned "official" and "scientific" (with graphs and everything!) and since they aren't engineers they take it as fact and sign up to it -- and when they see the resulting poor quality of the transcoder on a full encode, they think it's the best they can get (hell, they saw the graphs!).

I personally seek the best quality -- if it wasn't important to me I'd still be using VHS tapes. When I see these tests I turn my head and try to walk the other way, but I just can't take more than three steps without having to turn around and set the record straight.

[colemar]

Originally posted by Sir Didymus
Just want to add also that all transcoders keep intact the GOP structure of the source, while encoders have, in general, a different one. So you are comparing I frames to B or P frames, making not homogeneous the comparison.


I believe 700 frames are enough to cancel out the effect of comparing between I-P-B. I sampled with a 500 frame period, which is not a multiple of the GOP length (which should be 15).

[colemar]

Originally posted by jdobbs
I swore I wouldn't get involved in another of these ridiculous test reports, but here I am...

Why ridiculous?

The test was not telling that one program is the best in all circumstances. I even refrained to comment on the result of my test.
I described as accurately as I could how I did it, so that the reader can make its own judgement.

You can say the test tells little about the quality of a program with respect to final user satisfaction, but "ridicuolous" seems a bit excessive.

[colemar]

Originally posted by jdobbs
Then they see something like this that looks so damned "official" and "scientific" (with graphs and everything!) and since they aren't engineers they take it as fact and sign up to it -- and when they see the resulting poor quality of the transcoder on a full encode, they think it's the best they can get (hell, they saw the graphs!).


I don't see anyone here praising my benchmark as rocket science.

I like DVD-RB (I'm a subscriber).

I think in this particular case the metric agrees with my eyes, and shows that CCE is not always the winner.

[TheSeeker]

Originally posted by jdobbs
When I see these tests I turn my head and try to walk the other way, but I just can't take more than three steps without having to turn around and set the record straight.

I know exactly how you feel about this. This topic is something akin to political discussions. Everyone has an opinion and most feel strongly enough to voice that opinion. But, fortunately there are plenty of video professionals here to straighten out some of the newbies and temper the deluge of information.

[jdobbs]

@colemar

The point is (here I go again) -- you are making conclusions that others will take as fact that are based on flawed assumptions and an improperly constructed test.

Please don't misunderstand. I LIKE testing and comparing output of different encoders and I don't want to get into a war of words or start a flame. But, I want the facts to be presented accurately. I repeat: You CAN'T do a comparison of a transcoder to an encoder using SSIM. Truthfully you really can't accurately even do a comparison between two transcoders or two encoders using SSIM.

You are correct, CCE isn't always the winner -- but it is the winner almost always. So unless you want to do a time consuming comparison to decide which direction is the way to go every time you do a backup, you're better off just selecting the method that produces the best quality most often.

I want to be very clear here also. It isn't just CCE, QuEnc does better almost always, and HC does better almost always, TMPGEnc, ProCoder... The difference is between within-the-compressed-domain adjustments (transcoders) and true MPEG-2 encoders.

The only real test of motion video is in the viewing.

[ADDED] I also want to make clear the I am NOT slamming DVD Shrink. It is a fantastic program written exceptionaly well. It is rich in features, and is solid as a rock. That's one of the reasons you see so many people doing preprocessing with Shrink. They can work Shrink's magic on the source material -- and CCE's magic (Big 3 or DVD Rebuilder) for the quality.

[colemar]

Originally posted by jdobbs
The only real test of motion video is in the viewing.

Yes, but a subjective test should be a blind test with a control group. If not, the observer may be biased, and tends to select a particular copy based on some assumptions.
Frankly I don't have enough time to set up such a test.
And even a well conducted subjective test tells only the preference of a certain group of people.

I believe that in the test I made, the differences between the copies (let alone Intervideo) were nearly unnoticeable to the average observer.
In this situation, I think SSIM can discriminate enough to suggest what program performs better IN THIS PARTICULAR CASE.

[colemar]

Originally posted by Surf
Interesting even though I don't understand the techno details. I hope you can include Nero's Recode in your test. I have been assuming that it has an improvement over the dvdshrink....

Nero Recode 2 (2.2.6.9), Advanced analysis, High quality mode

Seems to perform a little better than DVD Shrink.

But only for this particular movie, and only in relation to SSIM metrics.

[dragongodz]

In this situation, I think SSIM can discriminate enough to suggest what program performs better IN THIS PARTICULAR CASE.
bu you have done a poll and not just said something like "these all looked very similar to my naked eye so the only difference i could find is the SSIM results". so whats the point ?

I even refrained to comment on the result of my test.
which is actually what you should be doing. asking people which line on the graph they like better means nothing.

fortunately there are plenty of video professionals here to straighten out some of the newbies and temper the deluge of information
i have to add.
i have also seen long time members(who like to think they are pros or since they have been encoding along time think they know just about everything) who also come out with some way off base statements. so its not just newbies that at times get things way wrong. thankfully this is in the minority for sure. ;)

[colemar]

Originally posted by dragongodz
bu you have done a poll and not just said something like "these all looked very similar to my naked eye so the only difference i could find is the SSIM results". so whats the point ?


I didn't intend to mean the poll as "which program do you think is better given my test results", but simply "which program do you like better".
It was obviously my fault to include a poll along with a such hot argument as quality measurement.

My reasoning is:

1. looking at the movie copies running doesn't show any noticeable difference
2. inspecting each still frame and comparing two copies along with the original shows noticeable differences
3. it is likely that these differences do have a meaning, that is, the copy that has frames more similar to the original should (perhaps) look better when running on a high end equipment and screened by a trained eye
4. but inspecting many (700) frames one by one is impractical
5. SSIM does show differences, and it agrees with eye inspection on a few random selected frames; I mean, if you inspect, say, frame 235 looking back and forth over the sequence copy1(235)-original(235)-copy2(235) and find that for example copy2 is more similar to the original (and better looking, when the original was well done), you will find that copy2(235) got a SSIM higher than copy1(235)
6. then it is very likely that SSIM can be a useful tool to automate the work described at 4
7. I'd bet though that even PSNR could be used in this situation

I understand the weakest point here is 3.

[dragongodz]

I didn't intend to mean the poll as "which program do you think is better given my test results", but simply "which program do you like better".
then that is the type of poll you should have done. then you could have said which you prefer and why(without graphs :)).

now to answer your points
1. since thats how you watch it thats what matters.
2. you do not watch movies as single frames(if you know what i mean). so single frames are fine to illustrate something you see during normal play but as means of comparison are very dubious.
3. thats a big perhaps and the results may not be inline with human results, making any meaning suspect. a metric measure will never beat or be as important as the human eye.
4. and pointless.
5. it agrees with single frame viewing, but again nobody watches a movie that way.
6. read my answer for 4. :D
7. and its been shown before how PSNR especially can be fooled to give a higher rating to worse looking footage.

[jdobbs]

There are lots of weak points.

1. What percentage of full length movies will fit on a DVD-5 at 73%??? Why didn't you try it at, say, 55-60% which in my experience is much more common? Especially for full movie backups with extras and menus which is how the vast majority of people back up their DVDs.

2. What bitrate was used for the original encoding? Is this a "standard movie" How much action is there in "The Terminal" in comparison to other movies?

3. 700 frames is a ridiculously small number for a comparison. That's 28 seconds or so... how many movies have you watched that are 28 seconds long? Were there any high action scenes? How about still scenes?

I could go on-and-on. I'm not asking you to do any of this, though,... because as mentioned earlier in addition to being incorrectly structured, the test uses techniques that can't accurately compare the output anyway.

EXAMPLES: Here are the four movies I did overnight last night: 46.2%, 56.1%, 51.2%, 69.3% -- not one out of four of them was above 70%. Transcoders would die miserably if they have to meet these kinds of reductions. Blocks on top of blocks. But with CCE they came out fine.

[jdobbs]

For those interested you should look here (http://forum.doom9.org/showthread.php?s=&threadid=91418) for a more realistic test. Real people, viewing real video, not knowing which was which, and deciding which one looks best.

[markrb]

I really have no opinion on the testing method or the idealogy used.
However I do have to disagree a bit with the idea that most people do full DVD backups.
I will agree that most of the people that have little working knowledge of DVD backups do go for the full backup because it usually is simplier.

Personally everyone I know that has beyond a basic understanding prefers to go movie only. In my case this is for 2 basic reasons.

1. Smaller amount to backup = greater space to devote to movie.
2. Most extra's are watch once (if that) and never again.

In my case over 90% of backups need little or no compression.
I don't believe I have done a Shrink with lower then a 75% level ever.

There have been movies I decided not to backup due to their size and being forced to go 65% or lower for exceptionally long movies. In these rare cases, for now, I have put them on hold.

Personally I find that when minimal shrinking is done the picture is superior to that of even a CCE 4 pass VBR. Fully encoding a movie tends to introduce unwanted noise.

However at higher shrink levels I do agree that CCE can't be beat, but in my case I have found very few movies that qualify.

Mark

[jdobbs]

However I do have to disagree a bit with the idea that most people do full DVD backups. Movie only backups are a lot less common than you think. The population at large (not just those who post on video forums) would rather just clicks and copy, and have a real backup... The only reason movie-only even got popular at all was because the result from the transcoders was too terrible to watch -- and at the time they were the only choice beyond reauthoring (which many, including myself did). To the old SVCD/VCD crowd is was what they'd gotten used to anyway. But I'll admit that much of what I say is just my opinion.

[dragongodz]

The only reason movie-only even got popular at all was because the result from the transcoders was too terrible to watch -- and at the time they were the only choice beyond reauthoring
sorry jdobbs but that is wrong. if anything transcoders brought full dvd backup to the masses. the software that used an encoder, such as dvd2dvdr etc, were all movie only. infact how many, besides DVD-Rebuilder, are full dvd backup even now ?
yes some people went to the trouble to completely reauthor dvds but the majority didnt because it was more hassle than it was worth for a plain movie dvd.

To the old SVCD/VCD crowd is was what they'd gotten used to anyway
yes exactly,vcd/svcd software was movie only and when the programmers moved to dvd backups they followed that through because it was the simplest thing to do. thus movie only became popular before transcoders were ever mentioned or released and it had nothing to do with them.

[jdobbs]

@dragongodz

Wrong? You know I respect your opinion, but I don't think so. I have to disagree right back. I'm not talking about unique tools that float around the DivX scene -- but the ones that hit the mass market.

The only reason movie-only even got popular at all was because the result from the transcoders was too terrible to watchI meant "popular on transcoders"

infact how many, besides DVD-Rebuilder, are full dvd backup even now ? All the transcoders do: Shrink, InstantCopy, DVD95Copy, DVDXCopy, DVD2One; because it is easy when you only make simple changes to the stream. In fact -- when DVD2One came out (the FIRST widely known transcoder) I remember the biggest complaint from most users was the fact that it wouldn't do complete disc backups. It got added within a month or so. People would have always done complete movie backups with the encoder packages too --> but it was too hard to do. That's why I wrote Rebuilder. Before then you had to reauthor with Scenarist or one of the other high-end packages. That was a real pain (even with the bunches of tools written just so you could).

I guess what I'm saying is: The quality folks who used encoders didn't make movie-only because it was their preference. They made movie-only because it was the only choice.

But... that's just me talking.

[dragongodz]

All the transcoders do:
you need to take the whole statement which clarifies the question
the software that used an encoder, such as dvd2dvdr etc, were all movie only. infact how many, besides DVD-Rebuilder, are full dvd backup even now ?
so the question wasnt how many transcoders do whole dvd but how many programs that use encoders do whole dvd backup even now. :)

I guess what I'm saying is: The quality folks who used encoders didn't make movie-only because it was their preference. They made movie-only because it was the only choice.
yes thats also what i said with
the software that used an encoder, such as dvd2dvdr etc, were all movie only.
and
yes exactly,vcd/svcd software was movie only and when the programmers moved to dvd backups they followed that through because it was the simplest thing to do.
so actually we pretty much agree there. :D
what i was saying is all this happened before transcoders were released. so people were doing movie only backups, not with much other choice, before transcoders so its not really fair to blame transcoders for it.

Wrong? You know I respect your opinion, but I don't think so. I have to disagree right back.
of course, no hassles to me. ;)

ummm isnt this all getting a bit off topic ? :D

[EDIT]
DOH! you edited while i was typing. :)
I meant "popular on transcoders"
sorry it was not clear that that is what you meant though.

[Stive]

Now here's a stupid question from a noob, is there enough data than can be analyzed by a program (perhaps something like dvd-rb) that would suggest this particular movie would be better encoded versus transcoded...and therefore, you select that method (rejig/shrink...whatever) or (CCE/Quenc/HC...whatever) and go your merry way? Yeah, I know, dream on.:D

[dragongodz]

is there enough data than can be analyzed by a program (perhaps something like dvd-rb) that would suggest this particular movie would be better encoded versus transcoded...and therefore, you select that method (rejig/shrink...whatever) or (CCE/Quenc/HC...whatever) and go your merry way? Yeah, I know, dream on
well this is really getting off topic and should have been a seperate post IMHO. but to answer your question actually yes you probably could. for example if the bitrate of the original footage is high and the reduction amount is small then a transcoder is likly to do a fine job the majority of the time.

[MackemX]

I've done a few polls over time regarding movie-only/full disc backups and it was always around the 50/50 mark with movie-only normally edging it

now with all the stripping tools available from blanking with DVDShrink or using more precise tools like DVDRemake/Vobblanker, people are removing unwanted video that normally wastes precious space. This obviously means a few movie-only people are now doing full backups as they can strip out the unwanted crap and keep what they want but still get a decent quality backup

as I always say, most who are that bothered about quality will probably do their own tests anyway. Sometimes even the movie determined which tool I used depending on how I rated the actual DVD. I haven't even watched some in my collection so why bother wasting time fussing?

Each to his own as the factors in the backup process are far too great to make a true comparision :)

[ankit]

Originally posted by dragongodz
well this is really getting off topic and should have been a seperate post IMHO. but to answer your question actually yes you probably could. for example if the bitrate of the original footage is high and the reduction amount is small then a transcoder is likly to do a fine job the majority of the time.

What would be the threshold on the movie bitrate and compression ratios to get a good transcoded result with shrink? I _always_ compress movie-only (if I want the extras, I split it to a 2-disc backup), and most of the time the compression ratios are in the range of 70%-90%. What would be the minimum target bitrate that would still have excellent results with a transcoder?

thanks!

[dragongodz]

What would be the threshold on the movie bitrate and compression ratios to get a good transcoded result with shrink?
hmm well how long is a piece of string ? :)
same answer, i dont know, it depends. it depends on the type of footage, the quality of the footage etc etc etc.
since all movies are different there is just no way to give a threshold.

[jdobbs]

Just a suggestion. This is getting off topic for this thread. You may want to think about starting another one.

[dragongodz]

well i did say it was off topic earlier. maybe a nice mod will split it ? :D

[TomBrooklyn]

Originally posted by ankit
What would be the threshold on the movie bitrate and compression ratios to get a good transcoded result with shrink? I like the results using Shrink down to about 65%. I usually compress menus and extras and trailers and stuff to the max to try to get the main movie compression as low as possible. I have some backups that had to go to around 55% on the main movie/video with Shrink and the quality is noticably reduced. Not terrible, but very noticable, and I didn't care for it. I will split anything I care much about that goes below 65% now.

I haven't tried Rebuilder yet because I haven't figured out how to use it. The Guide for it is pretty technical with a lot of acronyms and technical terms that make it not particularly friendly to someone like myself who is not an avid hobbiest, just someone who wants to backup DVDs with a minimum of fuss. If it can yield pretty good quality down to the 55% compression range that would relieve the need to split a lot of originals into two disks.

[blutach]

@TomBrooklyn

This (http://forum.digital-digest.com/showthread.php?s=&threadid=45129) is a very simple thread on compression ratio.

Regards

[TomBrooklyn]

That's interesting blutach. Most of the movies or videos I've backed up so far must have about the same bitrate as the quality seems to pretty consistantly drop off for me below 65% in DVD Shrink. I haven't backed up that many disks though, so my sample size is really way too small to draw a widespread conclusion, I suppose.

[mfsav2]

Can you test the latest clonedvd?

Mf

[petermarina]

Originally posted by mfsav2
Can you test the latest clonedvd?

Mf

Mf, I did a quick comparison between DvD-Shrink and CloneDVD2 a few weeks ago. At low compressions (85%) with recent and well encoded DVDs, clone is faster and as good as Shrink. But on an older DVD that was encoded with older encoder technology, it was just terrible. If I remember correctly it was at 67% compression more or less and the result was not watchable. By this I mean not worth wasting a 0.50$ DVD blank. I personaly stick with DVDShrink and try with DVD-Rebuilder/HC14 on older DVDs that require alot of shrinking.

At the time, I was so impressed with CloneDVD 2.8 I was about to buy it after the free trial ended. I did the test just in time and I am now 100% behind DVD-Shrink.

Hope it helps,

p.

[elizerrojas]

just stay away from this mess people and do what i do.


at 20% or less (100-80)compression ------- use shrink or recode2

at more than 20%or more(100-81)compression-----dvd RB/HC

but then again that is just my opinion.

P:s: with HC's quality+free, there is no need to buy CCE.

see you.

[jdobbs]

You should never make a decision based upon percentages... they are meaningless.

Which would you rather have on the same movie, a 70% reduction on an original that was at done at 5Mbs or an 90% reduction on one that was encoded at 3Mbs?

Of course I guess I'm an easy target for accusations of bias... but if an encoder technique always guarantees the best results possible, and a transcode only works sometimes (and you really don't know until it has already been done once) -- why not just always use the encoder? You're not saving time when you have to redo half the movies you start...

I just do all movies with a good encoder while I'm sleeping.