Detelecine/Decomb/Deinterlace on Futurama

[randybruder]

Yes, using the combination of Detelecine + Decomb on the Futurama DVDs is a large improvement over just using one of the Deinterlace options, which is an even larger improvement over encoding it straight...but my output is still iterlaced at some points. Only when characters in the show are small (i.e. off in the distance in the shot) they are still heavily interlaced, as well as the random interlaced frame.

Does this happed to anyone else? (Does anyone know the better way to encode it?)

_randy

[jbrjake]

...so change the decomb parameters.

[pudge]

I have tried dozens of different options for decomb settings, in both 0.9.3 and 0.9.4. Nothing seems to help. I gradually drop the motion/spatial detection, gradually drop the number of pixels ... I sometimes get some marginal improvement, but it's not long before I get interlacing artifacts.

Bleah.

I'd love some help here.

[jbrjake]

I don't know what number of pixels you're dropping. But if your problem is interlacing in tiny spots when characters are drawn with perspective at a distance, wouldn't it make sense to reduce the block dimensions and block threshold so that you're *looking* at tiny spots for combing?

[Starhawk]

I tried encoding the pilot for Futurama a few months ago, and I finally decided there was nothing I could do and chalked it up to just being the way the disc was encoded. Even using the slowest of slow decomb settings, combing was still noticeable.

But I am by no means an expert on the decombing options in Handbrake, so Im sure there is an answer for someone who knows more about it than I do.

[Deleted User 11865]

Even using the slowest of slow decomb settings, combing was still noticeable.

Decomb can be set to either Off, Custom or Default.

[Starhawk]

Sorry, I should have been more detailed but couldnt remember the custom decomb settings. By "slowest of the slow" I meant I tried many of the custom decomb combos including Mode 25 (EEDI2->yadif->mcdeint).

[jbrjake]

That makes no sense.

EEDI2 is an interpolator. All it does is filter interlaced frames. What bearing does using the "slowest of the slow" interpolators for the deinterlacing segment of decomb have on the combing detection segment? How is running over the detected frames more slowly going to affect which frames are detected in the first place?

Im sure there is an answer for someone who knows more about it than I do.

Totally at a loss for why I'm being ignored here. Just like i said above, if you want improvements when combing is only visible in small parts of the frame, you need to decrease the size of sections that are checked for combing, and decrease the threshold for seeing any of those sections as sufficiently combed to filter the frame.

[Starhawk]

EEDI2 is an interpolator. All it does is filter interlaced frames. What bearing does using the "slowest of the slow" interpolators for the deinterlacing segment of decomb have on the combing detection segment? How is running over the detected frames more slowly going to affect which frames are detected in the first place?

I have no idea. I've always noticed that the slower decombing methods were usually considered better and from a previous thread it seemed that Mode 25 was highly regarded and very slow (1-3fps for me). I quickly replied to the thread about something I did months ago and it seemed the best way to describe it.

Totally at a loss for why I'm being ignored here. Just like i said above, if you want improvements when combing is only visible in small parts of the frame, you need to decrease the size of sections that are checked for combing, and decrease the threshold for seeing any of those sections as sufficiently combed to filter the frame.

I'll certainly study up on how to change this and try it on Futurama when I have time.

[scooterbaga]

Totally at a loss for why I'm being ignored here. Just like i said above, if you want improvements when combing is only visible in small parts of the frame, you need to decrease the size of sections that are checked for combing, and decrease the threshold for seeing any of those sections as sufficiently combed to filter the frame.

Hello,

I'm having trouble with the default decomb, and I'm hoping for some sort of 'high quality' parameter set.

Default decomb:



How does one actually implement a fix for this? i.e. What do you type into the text field that pops up when you choose custom decomb?

I'm a little confused as to why no one ever posts their decomb parameters... (I'm assuming that the 7 numbers separated by colons is what's typed in to the custom decomb field.)

Thanks.

[mduell]

Totally at a loss for why I'm being ignored here. Just like i said above, if you want improvements when combing is only visible in small parts of the frame, you need to decrease the size of sections that are checked for combing, and decrease the threshold for seeing any of those sections as sufficiently combed to filter the frame.

I'm having trouble with the default decomb, and I'm hoping for some sort of 'high quality' parameter set.

Default decomb:

How does one actually implement a fix for this? i.e. What do you type into the text field that pops up when you choose custom decomb?

I'm a little confused as to why no one ever posts their decomb parameters... (I'm assuming that the 7 numbers separated by colons is what's typed in to the custom decomb field.)

Thanks.

And you decided to flail in the forums instead of reading the documentation because?

[jbrjake]

Note that page isn't updated for 0.9.4, and should only be used for learning the decomb parameters relevant to increasing detection sensitivity. See the top of libhb/decomb.c for the current parameter layout and defaults.

[makoffee]

Here's my answer:

custom decomb: 9:2:2:3:6:3:3
custom detelecine: 0:1:4:4:0:0
denoise: Strong
deblock: 5
picture cropping: Custom 0/0/8/6
anamorphic: strict
CQ: RF16
advanced string: cabac=0:ref=3:me=umh:b-adapt=2:weightb=0:weightp=0

It's not perfect, but after spending way to much time testing (over and over), I've been able to get the telecining and interlacing under control using those settings. The truth is the Futurama DVDs are a mess. Some scenes have telecining in fast motion, and other scenes are 100% interlaced. (such as major scene changes) To get a nice looking progressive image you need to set the filters up to be sensitive to both of these scenarios. I found that default decombing block (16x16) was way to large to catch some of the more subtle decombing artifacts such as mouth movements or Mr Panucci's hands in the scene above.

Like I said these settings aren't perfect but I've been able to get near progressive scan quality filtering with these, if you have any suggestions on how I could get the filters looking better please let me know.

[mduell]

denoise: Strong

Yow!

[Lifeisabeach]

I've been trying to get the "perfect" Futurama rips myself and have tried a number of options from various recommendations here and elsewhere in the forums. The last set of "recommendations" posted here quite frankly made my eyes bleed when I saw the results. I'm going to post what I've felt works best and why in the hopes it helps anyone else out. To begin with, I started with the Apple TV 2 preset, added various tweaks from other recommendations in various threads for animation, and made minor changes to get the following:

Detelecine: Default
Deinterlace/Decomb: Neither one.
Framerate: 23.976 (NTSC Film)
Anamorphic: Strict (no cropping)
Advanced x264 options: deblock=1,1:psy-rd=0.4,0.00:aq-strength=0.6:bframes=5:ref=6:b-adapt=2:me=umh:subq=9

(EDIT: almost forgot… Constant quality set to RF 16.)

The advanced x264 options I'm using were taken from a variety of threads with tunes for animation. The end results are noticeably improved over the plain vanilla ATV2 presets. The encoding times aren't unreasonable and the file sizes, at 200-220 MB vs 1 GB sources, are acceptable.

I used Strict Anamorphic because Loose tends to mess a bit with the dimensions on DVD rips. From what I understand, the encoder works best when the dimensions divide cleanly by 16, so I also left in the black side bars. They are trival in size; would be mimicked on-screen no matter where they are viewed anyway; and add negligible size to the file, so I simply saw no compelling reason to crop them since doing so would mean dimensions that didn't divide cleanly by 16.

Detelecining is a no brainer. Even the DVDs themselves show extensive telecining artifacts. Removing the duplicate frames restores the material down to how it was originally produced for a much cleaner appearance.

Deinterlace was a no-go, as recommended in the guide. Using it makes the videos look hideous. But despite numerous recommendations to use Decomb, I actually found that it didn't help for any of the more obvious remnant combing artifacts. Worse, it actually degraded some scenes that featured fast motion but otherwise had no combing artifacts to begin with. It's not as prevalent with the defaults, but I tried dozens of more aggressive tweaks to control the remnant combing artifacts and of those that worked in that respect, they also introduced interlacing artifacts where there was no need. The settings "recommended" a couple posts earlier are a good example of what I mean. The opening credits are butt-ugly all the way through. When it came right down to it, decombing meant trading one kind of ugly for another, so I left it off and live with what I have. It's probably worth mentioning that I have a 50" screen, so I may be a bit more sensitive to these artifacts than others with smaller screens may be. Still, that makes it that much more important for me to be particular about my end results.

As for the framerate, normally I would side with using VFR, but I have a Pioneer Kuro plasma set that normally runs at 60 Hz. It has an advanced option that, if enabled, looks for 24 fps material and then switches over to 72 Hz for a clean 3:3 pulldown. In one of my earlier test encodes, there was a noticeable amount of stutter or judder that went away when I switched to the fixed fps encoding. Since the material was originally produced for 24 fps and detelecining should be restoring it back to that framerate anyway, locking it in seemed reasonable and it certainly looks better for me. I suspect though that if I had a 120 Hz LCD TV, I'd be fine with VFR.

Anywho, I hope this proves of help to anyone else looking to get the best out of their Futurama rips. Unfortunately we are working with really badly produced DVD releases and unless Fox decides to remaster and re-release them, this is as good as it gets.