Instant Copy looks better than CCE

[YaoMing11]

Does it matter much if PowerDVD snapshots look sharper with IC8 encoded movies over CCE/DVD-RB encoded movies? I've tested 4 different movies, and took the same screenshot of the same scene to compare. CCE looks more blurry, while IC8 looks sharper. The compression of the movies range from 25-32%This dissapoints me because all this time I've been using CCE to do encodes, and now I see IC8 is noticably sharper when compared to it.

[Enots_]

It really isn't surprising that a transcoding application can out perform an encoder in terms of quality. The technical arguments goes along the following lines:
MPEG-encoding is a matter of choices: Gop-structure, motion vector length, inter/intra encoding, quantscale, ... This makes hundreds of billions of possible permutations for the same input material and finding the optimal solution is virtually impossible. Thus encoders search for a good solution.

Transcoders start out with an optimal solution: All the possible to generate quality is stored within the MPEG2-decisions made on the original DVD. What a transcoder such as instantcopy does is alter this set of optimal decisions marginally and not very surprisingly as long as you only marginally deviate from the given optimum your quality is very good. What I here mean by marginally is that only a small subset of decisions are reviewed by the transcoder (quant scale and coring in particular) where as motion vectors, inter/intra decision and so forth are all left alone.

The keyword is "marginally". As soon as the bitrates of the original and the desired destination are further appart things change. Decisions in the original are no longer optimal in the destination. A typical example would be that if bitrate is very high you'd want to go with mostly intra-blocks (Many different small advantages: independant of quantization errors in reference frames, faster to decode, ... - broadcast people use I-only 50+ Mbit/s MPEG2!) once bitrate goes lower you need to save bits and an interblock becomes the optimal choice. When bitrate nears the lowerlimits of what is possible entirely skipping the block(copying it from the frame before it) becomes the optimal solution. Transcoders usually does not review this set of decisions and thus will fail miserably when the distance is too big. An encoder would make this decision again and has much better odds of pickign somehting that'll look good.

And besides - InstantCopy isn't half bad....

[krkdnose3]

I did a bunch of tests recently myself using Rebuilder/CCE, Instant Copy 8, DVDCopy 2, and Shrink.

On dvds that were heavily compressed, Rebuilder/CCE and Instant Copy did the best job. The results were very close to each other and I was very impressed with Instant Copy. The other two didn't really give acceptable results...very blocky.

[ddlooping]

Hi all. :)

krkdnose3, what DVD Shrink settings did you use during your tests?

[jsquare]

Are you using the latest updated version of IC8?

Seems like they improved IC8 with update 8.3.0.4, maybe they're trying to keep up with Shrink 3.2 & DVD2ONE v1.5 which are very good compared to previous versions.

One thing they haven't fix and probably will never do is the under/over size problem, which could be very annoying sometimes.

[krkdnose3]

Originally posted by ddlooping
Hi all. :)

krkdnose3, what DVD Shrink settings did you use during your tests?

ddlooping,

The one dvd I did the most testing with was Joseph and the Amazing Technicolor Dreamcoat (NTSC). It had this one musical number where all these people were dancing around that seemed good for a comparison.

I did the full dvd backup with automatic compression and deep analysis. I did 4 different tries using the various AEC settings. The maximum smoothness gave the best result, but the people dancing around still looked more like pixelated blobs than people. I didn't really expect much because the compression was around 50%.

DVDcopy 2 was much worse than shrink, by the way.

I was very suprised at Instant Copy 8 (not the latest update by the way) because it looked so good. It did take a lot longer to do than Shrink, though.

If there's something else I should try, let me know because I really want Shrink to be better than the other ones. I used it exclusively until a few weeks ago when I started to wonder what else was out there. I love the interface and find it extremely user friendly.

[ddlooping]

Thanks for the reply, krkdnose3. :)

50% compression is indeed quite a lot for DVD Shrink to handle. ;)
IC8 result is quite surprising though.

[krkdnose3]

Originally posted by ddlooping
[B]50% compression is indeed quite a lot for DVD Shrink to handle. ;)

Yeah, I know :) I wanted to try an extreme case. I was really surprised by the IC8 result myself.

I still plan on using Shrink when more reasonable compression is required.

[infoscapeone]

Originally posted by ddlooping
50% compression is indeed quite a lot for DVD Shrink to handle. ;)
IC8 result is quite surprising though.

Why surprising? Since the first InstantCopy version a lot of people showed that InstantCopy is the highest quality - and takes longer to process. Personally I use only IC and don't bother if it takes an hour longer.

[ddlooping]

Surprising because I didn't think any transcoder would get good results at 50%.

v3.2 main goal was to equal or better Instant Copy.
We managed to achieve this during testing of the last few beta versions.
None of the beta-testers tried a title at 50% compression though.
At this point, all of us would split to two discs, use CCE, or wait for dual-layer. ;)

[jorel]

hy all!
@ infoscapeone
did you saw that thread and my comments about the screenshots posted?(my post #3 )
"...loose part of the image in the top and show more in the bottom comparing with the original and ...loose details on the top of the picture is easy but how instant copy found more details in the bottom from the original? how the borders stay with the same size but the picture inside change?"
please, take a look:

http://forum.doom9.org/showthread.php?threadid=81502

[infoscapeone]

Originally posted by jorel
hy all!
@ infoscapeone
did you saw that thread and my comments about the screenshots posted?(my post #3 )
"...loose part of the image in the top and show more in the bottom comparing with the original and ...loose details on the top of the picture is easy but how instant copy found more details in the bottom from the original? how the borders stay with the same size but the picture inside change?"
please, take a look:

http://forum.doom9.org/showthread.php?threadid=81502

I looked at the pictures right now. What is obvious to me is that the CCE encoded picture is far more blurry. From the other two versions (original and IC) I don't see a difference in detail sharpness. However, they seem to be different pictures. Especially when I look at the thumb (?) at the upper border, this seems to be a different size (1-2 frames off maybe).

But to come back at the original topic: Once the compression get's higher IC wins in quality while I agree that for lower compression Shrink and all the others are also OK. Personally I THINK that IC also delivers better quality (though not really visible) at less compression and so I use it always - I just don't mind the time it takes longer and I don't bother if the DVD is full or not. Personally for me only quality counts.

[jorel]

Originally posted by infoscapeone
I Personally for me only quality counts.

just like me,i think and search the same! :)
quality is a "group of factors" to the target and the link posted was only "informative".
i saw that IC results was very cool but CCE can give amazing results, depend how is used.

[djan]

Hi all,

I used IC8 for a time and I can tell that it gives very good quality. The difference with DVDShrink is that IC8 uses many pass for transcoding. Now to know if CCE does a better job, I don't know but the majority of CCE users will answer you yes !

[Enots_]

I think the technical arguments that a marginal change in bitrate would tend to favour InstantCopy over CCE. I presented those in an earier post in this thread. Vice verse when the change is no longer marginal then CCE will have the better cards. When a change no longer qualifies as marginal is subject to debate and likely to be heavily material dependant. I've not spend too much time with CCE so I'd hate to take a guess at where the point is where CCE will become better than InstantCopy, but I'll believe for an average movie that InstantCopy got the better cards for 20% reductions if the input material is not extreme. However doing a bitrate reduction from 5 to 1.5 CCE would probably win everytime. Anyway I heard lots of people arguing that CCE does better, but I've never seen any empirical evidence in that direction from them and those who has testet it on this board favoured InstantCopy.

At anyrate we're comparing two different things here and in my humble opinion both are the best for what they where created to do: Encoding or transcoding.

[jorel]

@ YaoMing11
do you have results encoding with CCE after solve problems in the script like posyed here? http://forum.doom9.org/showthread.php?threadid=81502&perpage=20&pagenumber=2 where waiting your results,you will be the "CCE withness users" about quality!

@ Enots_
welcome in forum and thanks...your first post and explanations are very cool! :) the "empirical evidence" will came from YaoMing11 results with CCE and if he don't get it(for some reason) i can do 3 encodes and post the resulting frames that you (and all) choose to compare with this arguments to compare quality:
same source--->target---> 1 dvd-r with
main movie(one audio,no subs) at 60% with dvdshrink
main movie(one audio,no subs) at 60% with IC8
main movie (one audio,no subs,no resizes)with CCE266(D2D-Roba)
i can post screenshots or cut part of the results!
i know why i choose CCE for that target to encode and want to show the reason ! :)

@ all
ideas are welcome

@ moderators
is possible to "mix" this thread here with : http://forum.doom9.org/showthread.php?threadid=81502&perpage=20&pagenumber=1
seamless tests and topics...thanks!
:)

[Enots_]

@Jorel
Though I am dying to see the results which I find very interesting I believe that this indeed is a very interesting question. Bitrate reductions are in their very nature very material depedant. Second the advantages and disadvantages of encoding over transcoding is bound to slowly change with the amount of bits you need to save. The question remains where which effect takes the upperhand. Thus to draw good inference on wether to choose one or the other method a large amounts of movies will have to be tested - and to make it even worse the testing of quality is in it's very nature subjective. I am most certainly too lazy to commit myself to a project of that nature ;) - I would kill for the results though....

However what would be the most interesting question is wether todays generation of transcoders could be a taken a step further to reevaluate basic decision types of the source that current ones do not. Especially I would assume that inter/intra/skipped blocklogic and motion vectors length decisions becomes suboptimal when the bitrate drops sufficiently. Reevaluating these could give some interesting improvements when many bits must be saved....

[TheSeeker]

@Jorel

Personally I think a better test would be to take two movies. One a very high motion lots of action very bright movie, the other low motion, dark, not much action. These I think would represent the two extremes of dvd content. Then run them through, 1:1 exact copies (not movie only) keep all the audio and subs. Do this just to tax the encoder/transcoder as much as possible. Then reduce them to almost the max amount, say like 50 or 55%. Then post results of some very telling scenes. And I say SCENES not screenshots as we are going to be watching motion video not still frame presentations when we watch the movie. Just save them as short vob clips or something like that. Then post the results. Personally I think we will find CCE doing much better on the high motion movie, with IC or shrink doing much better on the low motion. I think we can all agree that for low compression a transcoder is the way to go, not only for the speed but also they seem to have a less intrusive effect on the video, sometimes preserving more details then a total re encode will. Personally though I think that the output from IC seems a little blurred or not as sharp as say shrink with default sharp or max sharp aec settings. But that may be just me.

[jorel]

ok TheSeeker, good ideas!
one problem: i only use CCE with d2s that can do main movie only...believe,i don't like menus....my authored dvds(tmpgencdvdauthor) have chapters but the menu came only if i call it.....the movie auto-start,then what we can do? main movie only or another one can do the tests? trust me,i don't want to install lots to get the same! :o

[TheSeeker]

Well if you already have CCE installed you could get DVD Rebuilder (which doesnt even require an install) and do the cce backups with that. Or maybe find someone that already has it and have them do the cce portion of the test. Take a look at the dvd rebuilder forum on this site and see whats up.

[jorel]

all right, i will try DVD-RB!(i need to learn use it too, seems very cool) :)

[TheSeeker]

It is very cool. The only downfall right now is it cant handle seamless branching or multiple angle titles. But that should be remedied soon. The developer of Rebuilder has mentioned that support for these things will be coming soon.

[KeepWalking]

DVD-RB is the best for CCE conversions.

What the best program for screens capture?? I will do some tests to compare the quality of the encoders and transcoders, but I can't capture a same scene for it compares. I use power dvd, but but I do not obtain to accurately stop it in the same scene in each film.

Any help?:confused:

[jorel]

you can use virtualdubmod or virtualdubmpeg2 to get "perfect" frames!

[ddlooping]

Hi all. :)

KeepWalking, it is better not to use screen captures.
I'd suggest you create short clips (3~5sec), using DVD Shrink "Start/End" feature.
You can then make a compilation of these clips, with DVD Shrink Reauthor mode, and be able to compare them back-to-back.
You can even play them in slow-motion to emphasize the differences between each versions. ;)