:: DVD rebuilder & return of the king... dissapointing results...

@ddlooping:

Good - that's what I've been looking at so I haven't wasted any time


@everyone:

Here's my initial thoughts. First off, the original definitely suffers from encoding artifacts. Despite what many people think because of the generally high quality of DVD video, there are 2 problems that pop up in movies like this (ie long movie with high bitrate requirements). The first is that the bitrate isn't set as high as it could be if the movie was shorter or split across 2 discs. The second is that even the maximum possible bitrate on a DVD with a 5.1 AC3 track and no other audio isn't enough for some scenes. It's hard to catch it on a TV, because normal TVs aren't really displaying the original picture (they show a telecined version) and HDTVs have to upsample because their native resolution is higher than any DVD. Both situations cause some inaccuracies in the picture, even if the video is perfect, and if it's not perfect it's generally not much (if any) worse to the naked eye. This video stream happens to have a max bitrate of 7500kbps (according to the header), which is really kind of pitiful for a high motion DVD. I suspect that it's a result of making sure it fit on a single DVD-9, but whatever the reason, it ensures that some of what CCE sees as signal is actually noise.

The next thing I looked at was the effect of the conversion from YV12 to YUY2 and then back to YV12. Why is this important? Basically, MPEG video uses a compressed colorspace called YV12. It refers to the number of pixels that carry color information. One of the quirks of CCE, however, is that since it's designed specifically as a high end tool, it's not intended to use an MPEG as a source, and therefore it can't read YV12. In order to make it read YV12 encoded video, AviSynth converts it to YUY2, which is still a compressed colorspace, but with twice as many pixels carrying the color information. As you can probably guess, chroma upsampling (increasing the number of pixels with color information) can't be completely accurate. It's basically like converting an MP3 to a WAV file. It's really a series of educated guesses. The result will be something that's almost, but not quite, the same as it would have been without the conversion (like the source the original DVD was encoded from). To summarize, the original DVD is in YV12, AviSynth converts it to YUY2, and CCE changes it back to YV12 while encoding.

The reason I wanted to look at this is simple. It's a valid criticism of any CCE based method that it requires chroma upsampling. It obviously creates a certain amount of error before the encoder even does anything, and it's specific to CCE. As far as I know, every other major MPEG encoder can read YV12. It also slows AviSynth down, since AviSynth 2.5x works in YV12 natively, and the conversion process slows it down, therefore also slowing down CCE.

The difference between the original and the same file converted first to YUY2 and then to YV12, but not encoded to a lower bitrate, isn't noticeable to the naked eye, but AviSynth's SSIM filter does see a difference. When you add lowering the bitrate into the equation, it's not surprising that CCE would give you noise. You can probably eliminate some (or all) of the mosquito noise by changing the Quantization Characteristics (quality_prec) but I'd still recommend using some AviSynth filtering because the flaws are really just amplified from flaws in the source. I'd use at least UnDot, which may not actually help one bit, but unlike other denoisers it shouldn't do any damage either. If that doesn't help (and I suspect it wouldn't) I'd probably try adding Deen as well.

In reality, this is all speculative for me, because I would (and in fact did) split this movie across 2 discs. At a certain point I think it's unreasonable to expect good quality from any program, and since the original doesn't even have perfect quality, I don't expect CCE, Shrink, or anything else short of the hand of God (Don't just stand there gawking like you've never seen the hand of God before!) to do the job well. I'm also biased because I do my personal testing with a progressive scan DVD player connected to a 65" HDTV, and the defects stand out a lot better than on most people's setups (like mine) so my standards are a little higher.

 

Actually, that's not necessarily the case. Here's the difference. When you just send it to CCE and ask it to encode, you're telling it to make all the decisions for you on what details absolutely need to be kept. Unfortunately, while CCE is great at giving you the highest quality video given the necessary bits to do it with, it will also try to keep whatever details are possible when it doesn't have enough. Unfortunately that can cause problems with places it can't give you enough detail. I suppose I'd say I find CCE's results unpredictable if you don't give it enough bits to work with.

The other thing you need to remember is that you know something that CCE doesn't. You can look at the original and see the noise. CCE, on the other hand, is assuming that everything in the picture is signal, and that there is no noise. The big advantage to this is that you can try to determine what kind of noise there is and try to select an appropriate filter based on that determination. CCE also has built in filters that you can use, but they're not nearly as good as what AviSynth has.

Here's what I would do. Since you've identified the problem area(s) of the movie you can try encoding just that section with some filtering added. Here's some more specific detail from the AviSynth help file on Undot.

I don't like using anything else if I don't have to. It's hard to say how much it will help, but as I said, UnDot is very non-destructive. The only real exception to this, in my experience, is sources with noise added for effect, like Saving Private Ryan.

 

Well just for future reference (and hopefully the new results won't require it on ROTK) here's how you encode just one segment. The key is the Rebuilder.ecl file. It contains every ECL for every segment, so all you have to do is copy the appropriate lines out of it to a new file and name that file [whatever].ecl. In order to figure out what lines need copied, you need to know the name of the AVS file for the segment. Then you can use a text editor to find the first line with that name (minus the AVS extension) and you'll know where the section you need to copy is. I'll give you an example.

The first thing you need to do is copy this line from the beginning of the file:
; Cinema Craft Encoder Basic -- Encoder Control List
Make sure that's the first line in your new file. It's technically a comment (as you can see by the next line which is clearly not part of a standard ECL), but CCE reads it to make sure the ECL is for the correct version. Next, use the Find function in your text editor (I use vim but Notepad will work fine for this) to find the first occurence of the name of your AVS file, but without the AVS extension (ie V01000400001005 or whatever it is). Copy everything from the line above that one, which should say [item], to the one before the next line that says [item]. Make sure your new file has an extension of ECL and open it with CCE. That will load all the encoder settings that DVD-RB would use for that segment. Then click the encode button and let CCE do its thing.

It looks a little confusing reading it, so let me know if you have any questions.

Edit: Oh, and don't get my post yesterday wrong. I'm not trying to judge anyone else's preferences for backups, and I can understand wanting to keep things on one disc, but I still think you're all a bunch of whackjobs for the amount of compression you're willing to use on a disc Just kidding of course
_X_X_X_X_X_[small]Looks like I picked the wrong week to stop sniffing glue
DVD Rebuilder Guides: http://www.afterdawn.com/guides/archive/dvd_rebuilder_tutorial.cfm http://www.afterdawn.com/guides/archive/dvd_rebuilder_tutorial_advanced.cfm[/small]