DVD-RB Pro and Undot()?

[TECK]

When is recommended to use Undot()? Most DVDs are now very good.... so why use it? I believe that Undot() will help a little for a better compression.
The question is: Should I use it on every movie? Does it affect the movie quality in regular movies or only in episodes, like Sopranos for example?

UnDot is a simple median filter for removing dots, that is stray orphan pixels and mosquito noise. It basically just clips each pixel value to stay within min and max of its eight surrounding neighbors.

From Afterdawn.com:
http://www.afterdawn.com/guides/archive/dvd_rebuilder_tutorial_advanced_page_3.cfm

You understand now why I'm getting confused... some people say to use it, others NOT to...
I don't know anymore what to do... I used to use this filter in every movie done and I wonder if it was a good idea, after reading that in most cases this filter will not do much on modern DVDs, or damage the image...

Thanks guys.

[writersblock29]

@Teck

It really depends on the source. If you're having to resort to lower bitrates (say, 2.5K), it's a good consideration. If your source is old and/or poorly compressed to begin with (Miami Vice Season One, for example), it may also be a good choice. ON THE OTHER HAND, it may sometimes detract from the original: Some films have grainy appearances for "mood," such as Saving Private Ryan, and using any filters whatsoever is going to result in a slight-to-major distortion of your material. All filters distort -- that is, take away -- from the original's image. The real question isn't really "Why wouldn't I always want to use this?"... it's more a matter of when the distortion (or smoothing) caused by the filter(s) is less objectionable than the macroblocking that may occur from lower bitrates. I've read a lot of posts that recommend Undot for low-bitrate animation projects -- or meerly "cleaning up" the noise from old or poorly compressed sources. My own experiences with Undot has shown me that it's a tool... nothing else. No miracles.

{Edit} I really dig your skin for Rebuilder!

[TECK]

So basically, I should simply toss the filter and do the movies without. Most of the new movie releases don't need the filter, right?
I'm concerned with the quality of the movie... Also, how do you check the bitrates if they are not to low, as you mentioned?

I'm simply disappointed that I did most of my movies with Undot()... I'm sure it affected in a way the quality of the original.
I was simply following an advice from someone else and use Undot() unconditionally... until I started doing few movies without and notice better results... crap. My lost.
It's obvious the difference is minimal, but THERE IS a difference.

If anyone else can share their experiences about this issue, I appreciate that you post it here, so I learn more about it.
I would like to know if it's recommended, on a regular basis, to use this filter or not.

Thanks guys.

PS. Thanks for the skin compliment. :)

[jdobbs]

My opinion... never, ever use any filter unless you have a reason -- and then use if for a specific purpose based upon what you see on the original. Let's say for example you noticed your original copy of an old B/W film seems to have "speckles" that come and go -- you can clean it up with undot.

There are DVDs out there for which the authoring house just didn't take the time to clean up the source. That's where filters come in. One of the reasons AVISYNTH is included as a part of DVD-RB processing (it would be just as easy to just read the D2V) is because it has a filter written for just about anything.

[Rockas]

... and that's why I'll eneble filters test on the next release of RME... nothing like some good "own made" (lol) tests to get your eyes saying... "This is the best solution" :)

[Carpo]

so really then we shouldnt use filters unless the source is old and messy ?

after reading a few posts in here im confused as to how to back up a dvd now ;) - they say reading never hurt anyone (heheh)

so its best to do two passes (any more and its a waste of time exercise)
dont use filters unless you have too
matrix - always a good idear to use them or it a use your jugdement
qmatop (by sapstar bit old now and his encoder has advanced qmatop in it) it tweaks ecl file by looking at matix you use and using avs2avi and dctune - should you use that or once again is it a judgement call

HC or cce which to use - i norm use cce basic on 2 passes but from time to time (test to test) hc has given better results although when i use it i fell like i wasted about 40£

any and all advise is welcome on what looks to be simple (stupid) newbie questions

[jdobbs]

so really then we shouldnt use filters unless the source is old and messy ?

after reading a few posts in here im confused as to how to back up a dvd now ;) - they say reading never hurt anyone (heheh)

so its best to do two passes (any more and its a waste of time exercise)
dont use filters unless you have too
matrix - always a good idear to use them or it a use your jugdement
qmatop (by sapstar bit old now and his encoder has advanced qmatop in it) it tweaks ecl file by looking at matix you use and using avs2avi and dctune - should you use that or once again is it a judgement call

HC or cce which to use - i norm use cce basic on 2 passes but from time to time (test to test) hc has given better results although when i use it i fell like i wasted about 40£

any and all advise is welcome on what looks to be simple (stupid) newbie questions

- 2 passes are good for most -- 3 for long movies -- more might make you feel better, but won't really help the quality.
- Filters can do other things too, like resizing -- changing aspecte etc. But generally they are to clean up source problems.
- I think setting a specific matrix for each of the bitrate thresholds in DVD-RB Pro can improve quality.

[Carpo]

so basicaly if it aint broke dont fix it :D

i'll stick with 2 passes
dont really need filters on the sources i have - they are quite good (16:9 is good enuff for me) :D
i'll select the matix according to bitrate dvd-rb says
and not use qmatop

cheers jdobbs

[TECK]

Thanks guys, that answers to all my questions...
One little detail jdobbs. When you say "long movies", you mean the DVD size or it's lenght in hours... ?
A better example. First, if the DVD is 8GB, but the movie is 1h30min and it has several extras worth to keep, you would set it at 3 passes or 2?
Same question for episode DVDs... 8BG, no extras but 3 episodes, each 1h.

Also, more about bitrate settings and matrix. Can we have a tutorial? :)

Thanks guys.

[jdobbs]

The real measure is how many total frames are encoded in the entire DVD (including extras and movie). That usually can be measured in time -- but in NTSC it changes a lot depending upon whether it is 29.97fps, 23.976fps, or a hybrid of the two.

[Carpo]

if any of you get the road warriors dvd i would suggest undot on that - lots of old 80's clips on there that are messy

[writersblock29]

@Carpo

*Re: Undot on Road Warrior*

It's funny you mentioned that; I juuust did that one and used Undot to clean it up. Worked great!

[Carpo]

ooo i actually made a post that helped someone :D

[TheSeeker]

Thanks guys, that answers to all my questions...
One little detail jdobbs. When you say "long movies", you mean the DVD size or it's lenght in hours... ?
A better example. First, if the DVD is 8GB, but the movie is 1h30min and it has several extras worth to keep, you would set it at 3 passes or 2?
Same question for episode DVDs... 8BG, no extras but 3 episodes, each 1h.

Also, more about bitrate settings and matrix. Can we have a tutorial? :)

Thanks guys.

Heres what I do with the matrices.

For main movie I use the encoder default (So any segment with a bitrate higher than 3,000)

For anything lower than 3,000 (so 2,000 to 3,000) I use the QLB matrice good for medium bitrates.

For things lower than 2,000 and for extras I use the bach1 matrice.

[Rockas]

Also, more about bitrate settings and matrix. Can we have a tutorial?
About the matrices... read the posts from manono on "Rebuilder's Matrix Editor" thread :)
and read this thread:
http://forum.doom9.org/showthread.php?t=97193&highlight=return+king

keep it UP

[TECK]

One more important detail...
Most of the movies that I process with DVD-RB Pro, I resize them with Lanczos4Resize() to fit perfectly to my HDTV screen size.
Should I aply in that case the Undot() filter? Or CCE will take care of it and make it look perfect without any wierd defects due to resizing...

Thanks guys.

[TheSeeker]

One more important detail...
Most of the movies that I process with DVD-RB Pro, I resize them with Lanczos4Resize() to fit perfectly to my HDTV screen size.
Should I aply in that case the Undot() filter? Or CCE will take care of it and make it look perfect without any wierd defects due to resizing...

Thanks guys.

You distort the original aspect ratios? Sounds.... disruptive to me. Don't you sometimes gets some nasty squished faces and bodies doing that?

[writersblock29]

No kidding! Resizing 1.85:1 to 16:9! Is it out of concerns of saving space for encoding, or just for the sake of being able to point at it and say, "See? I did that." Not that there's anything wrong with the pride approach; I put my own signatures on my own work all the time for exactly that reason -- to make ME happy. If it's out of concern for encoding space... the improvement will only be marginal at best. Of course, if the black bars from letterboxing a 1.85:1 image annoy you, I guess you could argue that this is improvement enough.

[TheSeeker]

I suppose if the black bar are annoying its somewhat understandable. But, to me the squished or skinny people on the screen would be infinately more annoying.

[TECK]

Actually the quality of the output movie is impecable. I usually change a movie from 2:35 to 1:85, it will fit perfectly to my Toshiba 51" HDTV screen. Other formats, I don't touch them because they are OK.
Let me tell you the exact procedure so you get a better idea.

1. I rip 5 seconds of the main movie, a bright area where I can define well it's actual size.

2. With Fit2Disc, I calculate the new dimensions:

http://img356.imageshack.us/img356/2464/fit2disc10qr.th.gif (http://img356.imageshack.us/my.php?image=fit2disc10qr.gif)

In our example, the script will look like:
Lanczos4Resize(720,464,66,49,588,378)
AddBorders(0,8,0,8)

The bars will make the real aspect at 1.886 and that is perfect, because I did my 54 points convergence in a way to allow me more space behind the screen, due to some issues I had with ExpressVu 16:9 software dimensions in their receivers.

3. I encode the movie with DVD-RB Pro and CCE SP 2.70.
Here it is the result of the same scene:

http://img383.imageshack.us/img383/6163/fit2disc26at.th.gif (http://img383.imageshack.us/my.php?image=fit2disc26at.gif)

Now, notice how accurate was reproduced the text shadow around "September 1930", no pixelation or any other artifacts...
So that's why I wonder if it's recommended to do Undot().
For the very last movies, I did not used Undot()... actually the picture was a little better, in detriment of loosing some compression... but that's why we have the CCETargetSectors setting, right? :)

Let me know guys if you have a better way. Thanks.

[writersblock29]

I dunno... Even though you only have to trim a little from the sides, I'm too much of a purist to want to try it. I like the "as the director intended it to be seen" approach too much. But that solves wondering if you're watching squished images, at least.

[TheSeeker]

Yea no doubt. Your chopping off a world of visible screen on each side. The sides are important areas for directors to use to convey certain elements of feeling. For example, a director may put an actor at the extreme lower left hand side (the left hand side is the "weak" side) to convey that the actor is alone against all odd or he has some insurmountable task or he is in a weak position. With your end result, we may not even see an actors whole face if such a scene were to occur (where an actor is on the extreme edge of the screen). Too much lost for my tastes.

[Aquilonious]

- 2 passes are good for most -- 3 for long movies -- more might make you feel better, but won't really help the quality.
- Filters can do other things too, like resizing -- changing aspecte etc. But generally they are to clean up source problems.
- I think setting a specific matrix for each of the bitrate thresholds in DVD-RB Pro can improve quality.


I recently did a backup of a movie (movie + DC only) where the source matrix was the MPEG standard. But the bitrates DVD-RB gives back after Prepare phase are all above 5000.

In this case, would it be better to use the orginal matrix (MPEG std) or a better one like Fox Entertainment since the bitrates are high?

I'm wondering if it's just better to always extract and use the original matrix or to match the matrix according to DVD-RB Prepare pahse results.

p.s. I've tried RME but see negligible difference between the two.

[TuRiSOft]

...after Prepare phase are all above 5000.

In this case, would it be better to use the orginal matrix (MPEG std) or a better one like Fox Entertainment since the bitrates are high?

IMHO the source is affected by MPEG standard Matrix quantizers (used in the original encoding) and you'll never get better quality even when using matrices projected for higher br.
I think matrices can help when lowering br real much, say for example from 5500 to 3500 , and original matrix was projected for br above a given point (i.e. 4000 and higher) then we can try to change the matrix used for encoding according with dest br!!!

[feedback]

IMHO the source is affected by MPEG standard Matrix quantizers (used in the original encoding) and you'll never get better quality even when using matrices projected for higher br.
But quality is in the eye of the beholder.
I use a different Matrix all the time with High bitrate movies. For instance, the high bitrate of my "Band of Brothers" DVD set does not help the fact that the Director "Steven Spielberg" likes to use a very grainy effect in this movie, which I hate BTW. So, eventhough there is a high bitrate I use a different Matrix "BDVD or AVMAT6" among others to get rid of the grain. Therefore I produce a movie that is pleasing to MY eyes.

A high Bitrate alone does not a great movie make IMHO.

Regards,:)

P.S. I use the Rebuilder Matrix Editor all the time to check different Matrices of which I have many and since the filters were added I use those as well. The extract feature is nice in that it allows one to see the Matrix that was used in the making of the final movie.
I am amazed at all the work that goes into making a movie and upon final mastering they just throw up anything in the way of a Matrix and hope for the best.

[phædrus]

Why have all these tools if we can't "play director" a little bit?

I personally never liked the 2.35:1 aspect, and I can see why TECK wants to resize it and cut off the sides to make it 1.85. I never felt like anything was lost, when I tried a trick like that. Only one movie I can think of that I would not do that -- Forbidden Planet. But to each his own.

feedback doesn't like Spielberg grainy effects, so he can filter them out. why not?

With these signal processing algorithms, movie watching is more interactive. WE get to see the movie the way we want to. A director isn't God. Ridley Scot took Harrison Ford's narration out of Blade Runner in the current DVD release. For my money, he was dead wrong. I stick with the old laserdisc.

I'm surprised at the negative comment about UnDot. I use it all the time making DivX backups. The effect seems extremely mild to me. If you are compressing a DVD, you can't get something for nothing. If you keep more "dots" you will pay for it somewhere else. I only started using DVD-RB Pro about two weeks ago, and haven't bothered with filters yet, but for long movies, I will definitely use UnDot.

[jdobbs]

Hmmm... off topic, but I sure wish there was a DVD of the original theatrical release of Blade Runner. The "Director's Cut" ruined it.

My opinion is that you should be able to change the DVD any way you want. I don't expect Chrysler to start bitching if I decide to paint my car a different color or put in a different radio... it's my damn car.

[FlashBlade]

IMHO removing the grainy look of a specific movie is as wrong as cropping the image to make it fit the entire TV screen. I'm not sure many people would appreciate if you try to saturate the colors of the Mona Lisa because you think it's lifeless.
Visual look or aspect ratio are usually decided based on artistic decisions that ought not to be changed. Why not just re-edit the movie if you don't like some parts?
And you certainly cannot compare a movie to a car.

But all these points certainly go beyond technical aspects.

Hmmm... off topic, but I sure wish there was a DVD of the original theatrical release of Blade Runner. The "Director's Cut" ruined it.
I personally hate the original cut. The voice over was made to clarify points but gave a rather dull result (same thing for the "happy ending"). And no, Blade Runner is not a Film Noir...

[writersblock29]

Isn't this thread beginning to teeter on the brink of a flame war?

I think maybe we ought to simply say, "To each his own" and be done with it.

[Boulder]

IMHO removing the grainy look of a specific movie is as wrong as cropping the image to make it fit the entire TV screen.

UnDot really doesn't have that big effect on a video. You can check it yourself with a simple script which interleaves the original and the undotted frame. Also, re-encoding tends to make the grain "swim in the background", especially if you don't have enough bits to spare, and that's really annoying.

I personally hate the original cut. The voice over was made to clarify points but gave a rather dull result (same thing for the "happy ending"). And no, Blade Runner is not a Film Noir...
The original cut is actually the one Ridley Scott originally presented as his cut, this has been verified by producer Alan Ladd.
http://www.hometheaterforum.com/htforum/showthread.php?s=&threadid=192677
It seems that a certain Jerry Perenchio will make sure that it's going to be a cold day in hell before we see any true Special Editions of the movie. Which is really sad :(

[feedback]

IMHO removing the grainy look of a specific movie is as wrong as cropping the image to make it fit the entire TV screen. I'm not sure many people would appreciate if you try to saturate the colors of the Mona Lisa because you think it's lifeless.
Visual look or aspect ratio are usually decided based on artistic decisions that ought not to be changed. Why not just re-edit the movie if you don't like some parts?
And you certainly cannot compare a movie to a car.

But all these points certainly go beyond technical aspects.


I personally hate the original cut. The voice over was made to clarify points but gave a rather dull result (same thing for the "happy ending"). And no, Blade Runner is not a Film Noir...
So, we have a purist amongst us!

To use your analogy, I agree The Mona Lisa is art and I would do nothing to change it. However, Picasso's work looks like something my kids did IMO and I would have no problem changing Pablo's cubist work to something that was more appealing to me.

The director is not God and if I want to change a DVD to look the way I want it to look I will.
I will use Undot, Deen, AutoCrop and any and all filters, Matrices, or anything else that I wish to use on the DVD's that I own.

If I want to shoot skeet and use my DVD collection as the targets, that is my business.

[jdobbs]

Ok... let's all talk about something else. No sense arguing over something that in the end comes down to personal opinion.

[Aquilonious]

jdobbs:

I've tried Undot and it really doesn't appear to remove any significant (if even detectable) amount of grain.

Have you ever used the Pixedust (2) filter? Seems to be a better filter than most to reduce grain.

You can see the results of Socio's post here:

http://forum.doom9.org/showthread.php?t=85790&page=5&pp=20&highlight=noise+filter

I'd like to know where I can find this filter and how to use it in RB Pro since I'm not good at script writing.

[jdobbs]

Look on http://www.avisynth.org/warpenterprises/ it is a part of the Dust filter group.

[Boulder]

It's also an Avisynth 2.0x plugin which means you need to do some tricks to get it working with AVS 2.5x. There should be some threads explaining that, IIRC there's a function named LoadOldPlugins somewhere which automates the process.

[steptoe]

Without starting a war on what filter is best for what, there is already plenty of discussion on what filter can achieve this and that on the AVISynth forums, if you think undot() is not enough, I've had very good results on poor quality sources using a filter called RemoveDirt()

It has plenty of optiosn to play with to get the best out of it, and removes a lot more than undot() and deen() do with doing to much damage


I had a very bad source that was badly scratched and terrible quality running two filter by teh same author without changinga ny settings made a huge difference to the final output, proving that filtering can and will make a difference BUT each source IS different and so settings must be tweaked for that particulat source

Just reading the forums on avisynth should prove just how powerful it is and what a good addition it is for DVD-RB, people using avisynth spend a lot of time and effort getting the best out of filters and do achive very impressive results, but this all takes time and effort. In the cases of some of the better but slower filters a lot of time. We are talking days here they are so slow, but for the best quality time is not a factor

[Boulder]

You'll have to be careful with RemoveDirt, it sometimes causes some objects to move jerkily. The latest version should attack this problem but I don't know if it really is trouble-free.

[DvPs]

hi

i am using this filters:
Undot()
asharp(1,3)
ConvertToYUY2()
TemporalSoften(1,6,4,15,2)

and after reading your posts i think, that this is too much?
would it be better to use no filter for movies with a bitrate above 3000?

and i have a question: do you use any filters for the extras?

and if you do so, how do you set filters for the extras and no filters for the main movies?
do you work with rb-optimizer after the prepare phase?

mfg DvP

[jdobbs]

My personal opinion -- filters are meant to be used as the exception rather than the rule. A filter should be implemented for a specific reason. If you are getting dots -- use undot(), if you are getting noise, use a denoising filter, etc... but in the majority of instances with backing up a DVD the best results are achieved with no filters.

[DvPs]

@jdobbs

is it a good idea to encode even the extras without a filter? because they have normaly a low bitrate.
i use the "half d1 and half space" option

and i have another question:
are you thinking about a other option in dvd rb pro: different filters for the main movie and for the extras.
this would be great.

mfg DvP

[Boulder]

hi

i am using this filters:
Undot()
asharp(1,3)
ConvertToYUY2()
TemporalSoften(1,6,4,15,2)

Sharpening and then smoothing is not a really good idea. Besides, aSharp tends to make edges fat as it uses unsharp masking.

For very light denoising, mostly for increasing compressibility a bit, I use RemoveGrain(mode=5,modeU=17).TemporalSoften(2,3,3,6,2). Another one to try would be LRemoveDust(17,1), depending on the amount of noise or sharp grain in the video.

Also, put ConverttoYUY2() as the very last item in the script for maximal performance.

[jdobbs]

are you thinking about a other option in dvd rb pro: different filters for the main movie and for the extras.Yes, I'm looking at that for v1.10

[J-Wo]

P.S. I use the Rebuilder Matrix Editor all the time to check different Matrices of which I have many and since the filters were added I use those as well. The extract feature is nice in that it allows one to see the Matrix that was used in the making of the final movie.
I am amazed at all the work that goes into making a movie and upon final mastering they just throw up anything in the way of a Matrix and hope for the best.
I just started to use RME and the world of matrix experimentation! I have to admit that jdobbs is right about filtering should be the exception and not the rule. For months I've been adding a combination of RemoveGrain(mode=5) and TemporalSoften(2,3,3,6,2) to do some soft filtering. They tend to smooth out flat backgrounds like walls behind people's faces which sometimes have "dancing grainy pixels", but at the same time the effect looks artificial and blotchy. I was of the belief that a little filtering can help gain more compression, thus allow the encoder more bitrate for any given scene. But I think I may be changing my opinion on that...

My question is how does one extract the original matrix? And if anyone has more links to other matrix threads here that would be greatly appreciated. thanks!

[manono]

Hi-

My question is how does one extract the original matrix?

Often it's "original matrices". I often find up to 3 matrices in a movie, and I sometimes find them switching dozens of times. It looks to me as if there's a main matrix, one higher bitrate matrix to help bring up the static and black scenes to the minimum bitrate set, and another lower bitrate matrix used during complex scenes when the bitrate might bump up against the max set. This is a feature I wish CCE had, the ability to switch to a lower bitrate matrix during complex scenes, to keep the average quant used from skyrocketing during those scenes, and perhaps producing blocks. Of course, sometimes you see only one matrix used for the entire movie.

How to get them? One way is to run the Vob(s) through DGIndex, turning on "Log Quant Matrices" in the Options Tab. The other is to use RME, hitting the "Extract" button, and then "Extract Matrices".

I've been adding a combination of RemoveGrain(mode=5) and TemporalSoften(2,3,3,6,2) to do some soft filtering.

I'm glad you're seeing the error of your ways. Although you will gain compressibility, using temporal filters completely destroys the grain structure, and makes the movie look pretty peculiar. Although I use RemoveGrain for low bitrate extras, I don't dare use it for a movie I want to do with best quality.

[Aquilonious]

Hi-

My question is how does one extract the original matrix?

Often it's "original matrices". I often find up to 3 matrices in a movie, and I sometimes find them switching dozens of times. It looks to me as if there's a main matrix, one higher bitrate matrix to help bring up the static and black scenes to the minimum bitrate set, and another lower bitrate matrix used during complex scenes when the bitrate might bump up against the max set. This is a feature I wish CCE had, the ability to switch to a lower bitrate matrix during complex scenes, to keep the average quant used from skyrocketing during those scenes, and perhaps producing blocks. Of course, sometimes you see only one matrix used for the entire movie.


Please pardon my ignorance manono, but doesn't Rebuilder adjust your matrix on "on the fly" by specifying which ones to use for Main, LB, VLB, and Matrix for Extras? Shouldn't this solve at least part of the problem of peaks and lows?

I typically set Main matrix according to the bitrate from Rebuilder's prepare mode. For LB & VLB I use the corresponding named matrices, and for Extras I use med_med (I use CCE 2.70).

I find that testing different matrices in RME doesn't seem to make much difference, unless I choose a matrix that parts far from the original. As you mentioned, it's easy to see several matrices used in one movie when you extract them from the various vob files. For movies with several matrices I typically shoot for the mean average. Not perfect by any means, but usually I get good results.

Is this a good general solution, or am I missing something in my equation?

Also, do you know how I would write a script to use Pixiedust on vob files (outputted from DVDFab) in Rebuilder?

I've tried loading the PixieDust (DustV5.dll) plugin directly and also after loading LoadPluginEx, but after I run Prepare mode I either get errors in Rebuilder's Preview/Edit (ex. "error in line 8" in red letters) or Rebuilder crashes entirely, the only option being to use Microsoft's error reporting utility.

Right now I'm using a combo of STMedian+MSharpen to degrain but I don't like the blurring caused by STM, even though I'm using lower than default values for the filter. MSharpen helps a bit but using these two filters together adds tremendously to the processing time. It took 342 minutes to process a 2 hr movie on my Athlon XP 3200, with nothing else running in the background. Without filtering it never takes longer than around 200 min, even for a 3 hr movie.

[manono]

Hi again Aquilonious-

We're talking about 2 different things (I think). Do you remember when DVD-RB had the ability to use Adaptive Quants ("Adaptive Q-Matrix Switching")? It's still available when you use a recent CCE on its own. In cases where the bitrate dropped so much that it couldn't keep to the minimum bitrate set, a higher bitrate matrix would be used (the numbers were cut in half, or even to quarters). So during these static, very dark, or black scenes, you might get a higher bitrate matrix used, depending on what you set for the minimum bitrate. I believe you are using a lower bitrate matrix for those low bitrate scenes (not such a good idea for the main movie, in my opinion). For example; say you have a typical bitrate for a main movie of 4-4500. And say there are some scenes with a static camera and people just sitting around talking, maybe in the evening with low lights. I think it would be easy for the bitrate to drop to 2-2500 during such scenes. Do you really want to use one of the low bitrate matrices for such scenes, and have the scene become softer and blurrier? I think, if anything, that you'd want to use a higher bitrate matrix for such scenes, as CCE with the AQ box checked might.

Some recent hardware encoders seem to have the ability to use lower bitrate matrices when you get above a certain bitrate. I'm more-or-less guessing here, as I have no way to figure which matrix is used in which scenes in the commercial DVDs (maybe I could cut out those scenes and check-lots of work, since I can't duplicate it myself anyway). But it seems to me that when the bitrate jumps above a predetermined point (8000, 8500, 9000, whatever, or maybe it's done by switching to the lower bitrate matrix when the ave. quant is about to go over a predetermined point), that those hardware encoders switch to using a lower bitrate matrix. This would be a very handy feature to have for CCE, because then, by switching to a lower bitrate matrix during complex scenes, it would keep the everage quant from jumping up quite as high as it often does during such scenes. This is something not possible to accomplish in CCE, even when using DVD-RB.

A lot of this is just gross speculation, as I know nothing about hardware encoders. I can get the matrices used, but I don't know when in the movie they're used, or really even under what conditions they're used.

I've tried loading the PixieDust (DustV5.dll) plugin directly and also after loading LoadPluginEx, but after I run Prepare mode I either get errors in Rebuilder's Preview/Edit (ex. "error in line 8" in red letters)

Sorry, but I don't know anything about Pixiedust except that you have to ConvertToYUY2 before you use it. Perhaps open the .avs in VDubMod and see if it gives you a clearer error message. This thread should help:

http://forum.doom9.org/showthread.php?t=85384&highlight=Pixiedust

[Aquilonious]

Hi again Aquilonious-
Sorry, but I don't know anything about Pixiedust except that you have to ConvertToYUY2 before you use it. Perhaps open the .avs in VDubMod and see if it gives you a clearer error message. This thread should help:

http://forum.doom9.org/showthread.php?t=85384&highlight=Pixiedust

Thanks for the link. I now have PixieDust working and use the following script in Rebuilder:

LoadPlugin("C:\Other\Plugins\LoadPluginEx.dll")
LoadPlugin("C:\Other\Plugins\DustV5.dll")
Converttoyuy2()
PixieDust(2)
Converttoyv12()

I was making two mistakes. One, I had Dust loaded in the Avisynth plugins folder. I needed to load it from a separate folder on my drive. Two, I was not converting color space values before and after PixieDust.

Whether I'm using STMedian or Dust, it slows down processing considerably--by about 400%. Groan. I really need to upgrade! :(

BTW, I've noticed that CCE's MPEG matrix (labeled med_high in RME) is different than the MPEG matrix that comes with Rebuilder. The non-Intra numbers are quite higher in the Rebuilder MPEG matrix than CCE's.

CCE Medium_High Non-Intra
16 16 16 16 16 16 16 16
16 16 16 16 16 16 16 16
16 16 16 16 16 16 16 16
16 16 16 16 16 16 16 16
16 16 16 16 16 16 16 16
16 16 16 16 16 16 16 16
16 16 16 16 16 16 16 16
16 16 16 16 16 16 16 16

Rebuilder's MPEG Non-Intra
16 17 18 19 20 21 22 23
17 18 19 20 21 22 23 24
18 19 20 21 22 23 24 25
19 20 21 22 23 24 26 27
20 21 22 23 25 26 27 28
21 22 23 24 26 27 28 30
22 23 24 26 27 28 30 31
23 24 25 27 28 30 31 33


Which of the two is actually the real MPEG standard number?

It appears that Rebuilder's MPEG matrix allows for higher compression.

[manono]

Hi-

I don't think I'd add a ConvertToYV12 at the end of the script, as I believe that DVD-RB is just going to (rightfully) convert it back to YUY2 before feeding it to CCE. This assumes that PixieDust is the last line in your script (no Undot, for example, coming after PixieDust). You can check one of the DVD-RB scripts to confirm (or disprove).

DVD-RB changes a couple of the matrix names slightly. The 2 low bitrate matrices have their names changed slightly (for clarity or simplicity, I assume) from what CCE calls them. Anyway, I'm encoding at the moment and can't check. But the one with all 16s is what I call the MPEG matrix, and what (I think) CCE calls the MPEG Standard matrix. You'll see both names used in the RME description. The one that goes up to 33 is CCE's Standard (and default) matrix. They're just names. As long as you know what you're getting, I don't guess it makes much difference. Not much help, eh? :)

[Boulder]

In CCE the one with all 16's is called MPEG standard. In HC, the built-in MPEG matrix is the other one mentioned here (CCE standard). The confusion might lie there.

[Aquilonious]

@manono -

I just pulled the script from the forum and removed the line referencing Unfilter (don't use unfilter). Since the link is referring to an older version of Rebuilder, perhaps the ConvertToYV12 function had not yet been implemented. In any regard, I removed it. But wow is PixieDust slowwww! Take a look:

[01:41:21] One Click encoding activated...
-----------------
[01:41:22] Phase I, PREPARATION started.
- CCE SP 2.70.2.1 encoder selected.
- AVS Filters are enabled.
- VTS_01: 2,155,000 sectors.
-- Scanning and writing .D2V & .AVS files
-- Processed 182,190 frames.
-- Building .AVS and .ECL files
- Reduction Level for DVD-5: 102.8%
- Overall Bitrate : 5,207/4,165Kbs
- Space for Video : 3,863,662KB
- HIGH/LOW/TYPICAL Bitrates: 8,264/3,381/4,165 Kbs
[01:44:14] Phase I, PREPARATION completed in 3 minutes.
[01:44:14] Phase II ENCODING started
- Creating M2V for VTS_01 segment 0
- Creating M2V for VTS_01 segment 1
- Creating M2V for VTS_01 segment 2
- Creating M2V for VTS_01 segment 3
- Creating M2V for VTS_01 segment 4
- Creating M2V for VTS_01 segment 5
- Creating M2V for VTS_01 segment 6
- Creating M2V for VTS_01 segment 7
- Creating M2V for VTS_01 segment 8
- Creating M2V for VTS_01 segment 9
- Creating M2V for VTS_01 segment 10
- Creating M2V for VTS_01 segment 11
- Creating M2V for VTS_01 segment 12
- Creating M2V for VTS_01 segment 13
- Creating M2V for VTS_01 segment 14
- Creating M2V for VTS_01 segment 15
- Creating M2V for VTS_01 segment 16
- Creating M2V for VTS_01 segment 17
- Creating M2V for VTS_01 segment 18
- Creating M2V for VTS_01 segment 19
- Creating M2V for VTS_01 segment 20
- Creating M2V for VTS_01 segment 21
- Creating M2V for VTS_01 segment 22
- Creating M2V for VTS_01 segment 23
- Creating M2V for VTS_01 segment 24
- Creating M2V for VTS_01 segment 25
- Creating M2V for VTS_01 segment 26
- Creating M2V for VTS_01 segment 27
- Creating M2V for VTS_01 segment 28
- Creating M2V for VTS_01 segment 29
- Creating M2V for VTS_01 segment 30
- Creating M2V for VTS_01 segment 31
- Creating M2V for VTS_01 segment 32
- Extracting STILLS for VTS_01 segment 33
- Creating M2V for VTS_01 segment 34
- Creating M2V for VTS_01 segment 35
[12:32:09] Phase II ENCODING completed in 648 minutes.
[12:32:10] Phase III, REBUILD started.
- Copying IFO, BUP, and unaltered files...
- Processing VTS_01
- Reading/processing TMAP table...
- Rebuilding seg 0 VOBID 1 CELLID 1
- Updating NAVPACKS for VOBID_01
- Rebuilding seg 1 VOBID 2 CELLID 1
- Rebuilding seg 2 VOBID 2 CELLID 2
- Rebuilding seg 3 VOBID 2 CELLID 3
- Rebuilding seg 4 VOBID 2 CELLID 4
- Rebuilding seg 5 VOBID 2 CELLID 5
- Rebuilding seg 6 VOBID 2 CELLID 6
- Rebuilding seg 7 VOBID 2 CELLID 7
- Rebuilding seg 8 VOBID 2 CELLID 8
- Rebuilding seg 9 VOBID 2 CELLID 9
- Rebuilding seg 10 VOBID 2 CELLID 10
- Rebuilding seg 11 VOBID 2 CELLID 11
- Rebuilding seg 12 VOBID 2 CELLID 12
- Rebuilding seg 13 VOBID 2 CELLID 13
- Rebuilding seg 14 VOBID 2 CELLID 14
- Rebuilding seg 15 VOBID 2 CELLID 15
- Rebuilding seg 16 VOBID 2 CELLID 16
- Rebuilding seg 17 VOBID 2 CELLID 17
- Rebuilding seg 18 VOBID 2 CELLID 18
- Rebuilding seg 19 VOBID 2 CELLID 19
- Rebuilding seg 20 VOBID 2 CELLID 20
- Rebuilding seg 21 VOBID 2 CELLID 21
- Rebuilding seg 22 VOBID 2 CELLID 22
- Rebuilding seg 23 VOBID 2 CELLID 23
- Rebuilding seg 24 VOBID 2 CELLID 24
- Rebuilding seg 25 VOBID 2 CELLID 25
- Rebuilding seg 26 VOBID 2 CELLID 26
- Rebuilding seg 27 VOBID 2 CELLID 27
- Rebuilding seg 28 VOBID 2 CELLID 28
- Rebuilding seg 29 VOBID 2 CELLID 29
- Rebuilding seg 30 VOBID 2 CELLID 30
- Rebuilding seg 31 VOBID 2 CELLID 31
- Rebuilding seg 32 VOBID 2 CELLID 32
- Updating NAVPACKS for VOBID_02
- Rebuilding seg 33 VOBID 3 CELLID 1
- Updating NAVPACKS for VOBID_03
- Rebuilding seg 34 VOBID 4 CELLID 1
- Updating NAVPACKS for VOBID_04
- Rebuilding seg 35 VOBID 5 CELLID 1
- Updating NAVPACKS for VOBID_05
- Updated VTS_C_ADT.
- Updated VTS_VOBU_ADMAP.
- Updated IFO: VTS_01_0.IFO
- Updating TMAP table...
- Correcting VTS Sectors...
[12:40:23] Phase III, REBUILD completed in 8 minutes.

Done.
[12:40:23] PREPARE/ENCODE/REBUILD completed in 659 min.

It took 659 659 MINUTES! to apply the Dust filter only. Ouch!


In CCE the one with all 16's is called MPEG standard. In HC, the built-in MPEG matrix is the other one mentioned here (CCE standard). The confusion might lie there.
@Boulder -

Which matrix is the genuine MPEG standard--the one with 16's (CCE) or the other with higher numbers? From what I've been able to tell, more movies use the CCE MPEG, so perhaps that is the standard. :confused:

[jdobbs]

Wow... I could have pulled out the film and erased the dust off of each individual frame faster. Hmm... undot().jdobbs() :eek:

[SpazzHH]

Which matrix is the genuine MPEG standard--the one with 16's (CCE) or the other with higher numbers? From what I've been able to tell, more movies use the CCE MPEG, so perhaps that is the standard. :confused:
Your assumption is correct. MPEG Standard is the one with all the 16's.

[Boulder]

But wow is PixieDust slowwww!
That's probably one reason why Didée developed the LRemoveDust function, which is a lot faster.

[Aquilonious]

That's probably one reason why Didée developed the LRemoveDust function, which is a lot faster.
Can you show or refer me to some sample LRemoveDust output, or a link of such output?

I've used STMedianfilter, which is a subset of RemoveGrain, but it blurs the output too much for my taste. I have to apply MSharpen afterwards to restore lost detail. Perhaps I'm not using the right values. I tried the default values (8,15,4,7) as well as (8,10,4,7).

You're absolutely right about PixieDust's slow performance. Right now I'm cleaning up Close Encounters and I started it at 4:30 this morning. It's now past 1pm and it's only 61.8% done, using only two CCE passes. Gosh, I think I need a rendering farm! But despite its glacial speed, PD does an excellent job of cleaning up video without much loss of detail, if any--at least in my limited experience.

BTW, what does SEETools do? From the RemoveGrain docs it looks like they're only applicable to interleaved sources. Does SSETools enhance performance (speed of operation) in any way?

[Boulder]

All the necessary information regarding LRemoveDust can be found in the official RemoveGrain thread by kassandro. Use a search and you'll find it.

He's got a forum over here : http://videoprocessing.11.forumer.com/ . You should ask any questions regarding his filters there as that's where he'll offer support.

[Aquilonious]

Wow... I could have pulled out the film and erased the dust off of each individual frame faster. Hmm... undot().jdobbs() :eek:
You mentioned in another thread that, "I think setting a specific matrix for each of the bitrate thresholds in DVD-RB Pro can improve quality."

Does this mean it would be best to set a specific matrix for Main, Low Bitrate, VLB, and Extras? Or should I just keep Rebuilder's default of "Same as Main Feature"?

For example, I typically set the Main Feature matrix after doing Prepare mode and looking at the bitrates. I set the LB matrix to LB, VLB to VLB, and for Extras I usually use med_med.

I'm a relative newbie to all this so I'm still a bit confused. :confused: