Reduction Level recommendations

[djan]

Hi all,

Can you give me some reduction level recommendations ? As to know when is it too low and the encoder will have difficulties ? 60% - 70% ? When to need to use denoising filter as Undot().Deen() ? Then I know if I should keep/low some extras or not. Per example at a reduction level of 50%, is it better to move out some extras or is there need to use filters ?

I know I should test by myself but I would also like to know your advises. Thx.

[feedback]

I don't really go by the reduction level, I go by the bitrate. I try to keep the bitrate above 3000 as I usually only do movie only because I like the highest quality on the main movie.

Using DVD Rebuilder Pro. I do use the blanking of some extras if I want to keep Menu and some extras. I use an ultra lowbitrate matrix on the extras that I do keep though.

Regards,:)

[writersblock29]

I'm with Feedback. Even though I know that bitrate in itself isn't anything to obsess over, it'll sometimes give you a pretty reliable rule of thumb. What I do with my disks is this: After decrypting, I'll open the project in either DVD Shrink or Recode so that I can see a breakdown of what's on the disk. The things I look for are 1) Total sound streams that can be removed and 2) total running time of all video on the disk. A disk with over 3 hours on it is usually going to involve a few tweaks somewhere. Sometimes, while the project's already loaded in 'Shrink or Recode, I'll leave the movie and extras untouched and shrink down the menus a bit. Many times, I'll run the output from this process through DVD Remake*, and get rid of as many things as I don't feel the need to keep (it's cool that Remake has an option to convert menus to still -- that'll save some serious space on certain disks).

*I need to play with Rebuilder's blanking feature... I just haven't yet because I've fallen into a routine since using the free version of Rebuilder, which didn't have this option.

Anyway, then I'll run a Prepare stage on my "final" project, and use the bitrate calculation as my rule of thumb. 3000 or above? Good times! 2800 thereabouts? I'll encode for awhile without adding anything, then abort the process and play the segments Rebuilder's worked on to see where I stand. Many times I'm still OK. 2500 or below... I either try filters, or re-evaluate what I really need in a backup disk. I'll usually just do a "movie only," since I rarely watch features and still have my original if the urge strikes me. Eposide disks are the exception -- and it really depends on the disk as to what seems to work. Dirty old footage (such as the Miami Vice disks) which looks like crap on the ORIGINAL definately benefits from running a smoothing filter or two to clear up the noise. Which ones, I'll leave to you, since flame wars are started as soon as a thread becomes a "what's best?" discussion. I sincerely hate having to tell you to dive into trial and error, understand -- your's is a good question! But like John Wayne said, "Don't call me Marion." After which -- when the cameras were rolling -- he said: "A man's gotta do what a man's gotta do."

[Rockas]

I'll encode for awhile without adding anything, then abort the process and play the segments Rebuilder's worked on to see where I stand.
You don't need to do that... just use the Rebuilder's Matrix Editor... It has a new feature that I added exactly to make previews of the final result :)

[djan]

I discovered something interresting using RB-Opt. The OPV testing tool for the Q value. It's a good way to test different settings changing avs file and CCE internal settings. I was guesting myself if the Q value was a good reference for quality. Meaning more the Q value is low, more the quality will be better. And upside wich value, the quality will be worse. I've read on the OPV readme file that a Q value greater than 40 will give bad quality. What do you think about ?

[robot1]

Q is a good measure of quality in first approximation.
Anyway, it's not perfect:
if you run OPV prediction for the same movie, one time without filters, and the second time with undot().deen() or other blurring filters, the second time Q will be lower of the first time... but you have to decide if the bluried movie is visually better of the un-blurried one (even if it has a higher Q, you could prefer its visual aspect).
Same thing applies with matrices: it's always good to check the result (and that's the usage of the "Check quality" button in RB-Opt)

[writersblock29]

@Rockas

*Matrix Editor*

As always, Rockas, you're a genius! Keep makin' us proud!

:p :p :p

[Rockas]

@Rockas

*Matrix Editor*

As always, Rockas, you're a genius! Keep makin' us proud!
:cool:
There are many "genius" on Doom9 forums (jdobbs, neuron2, Nic, dragongodz, robot1, manono and many others) I don't consider my self one of them :)

But thanks, anyway :D

[djan]

Q is a good measure of quality in first approximation.
Anyway, it's not perfect:
if you run OPV prediction for the same movie, one time without filters, and the second time with undot().deen() or other blurring filters, the second time Q will be lower of the first time... but you have to decide if the bluried movie is visually better of the un-blurried one (even if it has a higher Q, you could prefer its visual aspect).
Same thing applies with matrices: it's always good to check the result (and that's the usage of the "Check quality" button in RB-Opt)Ok but less is Q value, less there will be macroblocks/pixelisation. Isn't ?

[robot1]

Yes it is.
You have to check if it's better a few macroblocks, or a blurry result.

[robot1]

@Rockas
You're mixing "genius" and normal people...
;)

[Carpo]

maybe not about reduction levels persy but if you have a reduction level of 66% a bit rate of 3136/2616 (i see thats a new feature - orig bitrate -new bitrate - think it is anyway :)) and you have in your avs files ,idct=3,cpu=4 can you/do you need to have undot().deen() in there as well or will that just make it look nasty - btw amd using the 2800-3500 matix

[borgraf]

bit rate of 3136/2616 (i see thats a new feature - orig bitrate -new bitrate - think it is anyway :))the changelog explains it pretty well :)
- Modified the display of NTSC bitrates so the actual achieved bitrate (29.97fps) and the ECL file bitrate (23.976fps) are shown. In the past the lower (ECL) rate was displayed and prompted questions as to why it was lower than what might be hand-calculated. The ECL rate is used as part of the method to ensure support for hybrid sources and is readjusted as when necessary during REBUILD.

[Carpo]

at 6 am in the morn i find it hard to read - :)

[djan]

What's the advantage with encoding OPV instead VBR ? Just a time gain ? Will the quality be worse than VBR ?

[robot1]

There are different opinions about OPV vs VBR.
Anyway, I think the problem for OPV is the size prediction.
If the prediction is good (the size is similar to the one you get with VBR) quality is at least on pair.

Search for some posts from Sir Didymus in this same section: there are many comparisons.

[Edit]
Wrong spelling of the name of the great SD.
Sorry!

[djan]

Also, what Q value do you advise as optimal quality ? Meaning, under what Q value there is no need to go ? Is Q = 20 a good reference ? Or a value between 20-30 ? Will trying to get a value under 20 show any quality gain ?

[jdobbs]

Ok but less is Q value, less there will be macroblocks/pixelisation. Isn't ? The problem with filters is that quantization value you see after encoding measures the differences between the picture presented to the encoder and the output. A filter is applied before the picture goes to the encoder. So if I added a filter that blanked the picture out completely I could get a Q of 1... but is it really better?

[djan]

The problem with filters is that quantization value you see after encoding measures the differences between the picture presented to the encoder and the output. A filter is applied before the picture goes to the encoder. So if I added a filter that blanked the picture out completely I could get a Q of 1... but is it really better?Oh I see, so it's better to pass away filters for that quality test purpose. I always preferred to backup my movies as the original without filters. So the Q value test is a good way to know if the movie needs more bitrate or not.

[Carpo]

so which bitrate do u go by then for selecting matrices and stuff ?

[djan]

so which bitrate do u go by then for selecting matrices and stuff ?I don't use matrices.