How does MVBob compare to other bobbers?

[Chainmax]

I'm going to try BlendBob on an extremely difficult to IVTC source and want the absolute best bobber available. Speed is of no concern. Also, is the package in here:

http://forum.doom9.org/showpost.php?p=757259&postcount=7

The latest version?

[Guest]

I'm going to try BlendBob on an extremely difficult to IVTC source and want the absolute best bobber available. Speed is of no concern. Please read and follow forum rules, specifically, rule 12: do not ask what's best. Thank you.

[scharfis_brain]

mvbob is intended for true interlaced contents from video cameras.
Fieldblended or Telecined parts won't gain much compared to other deinterlacers like tdeint or kerneldeint. It even could be WORSE.

Please stick to the Fildmatchers & Decimaters avaiable. Such are - as you probably already know - TFM from Tritical, Decomb from Neuron2, and some others.

[DigitalDeviant]

I had to resort to Blendbob for some horribly telecined anime and got good results. I found tdient and (leak)kerneldeint to work the best for me. I tired MVBob too but didn't notice any gain in quality, but no loss either really.

[Revgen]

I prefer either EEDI2+TDeint or MVBob for bobbing video footage. MVbob is a little bit sharper looking to my eyes, also preserves slightly more details, but is much slower. EEDI2+TDeint is faster plus I can use TSP's SetMTmode filter for extra speed with my dual-core CPU.

I typically use EEDI2+Tdeint unless MVBob is really needed. The slight visual advantage that MVBob gives often isn't worth the slow speed IMHO.

[Zarxrax]

I find that TDeint(mode=1) works absolutely perfectly. Try it and see.

[Mug Funky]

Tdeint + blendbob works wonders.

blendbob doesn't give a vertical blur if the bob it's fed with is motion-adaptive :)

try:

tdeint(1,1,mtnmode=3)
blendbob()
decimate()

not many bizarrely telecined sources can get through that mix.

if you're a little more confident about finding good matches, add "tryweave=true" to tdeint somewhere. that means messing around with cthresh and MI as well, and wont give you much extra quality (as the default kernel mode plus mtnmode=3 are very well suited to blendbob), but will make things faster on a slightly less impossible source.

note, i usually IVTC with something much faster, but it fails more often and i have to fall back on something like the above.

[Alain2]

Thanks Mug Funky, you made my day with this simple script :)

I didn't know about blendbob and after your post I had a look in Leaks thread, this is doing exactely what i was thinking of scripting after unsuccessful attemps with tivtc, removegrain/temporalrepair, tdeint, etc. Thanks Leak!

[Uanime5]

When I tried to use this code I got the error message

Evaluation: system exception - Access Violation

It also says that this problem is with 'blendbob()'. I tried downloading the latest blendbob.dll but this did not help. How do I resolve this problem?

[Chainmax]

neuron2: sorry, title changed :o.

scharfis_brain, Revgen: the idea is to use this in conjunction with BlendBob in order to do IVTC. None of the automatic telecine methods I tried (SmartDecimate, TIVTC, Decomb) gave satisfactory results on this source.

Mug Funky: my computer is a poor little Sempron 2200+ with 256MB Ram, could you please download this sample capture (http://www.31012.com/~azulftp/X-Men%20-%201x01%20-%20IVTC%2001%20(HUFF).01.avi), try both your suggestion and MVBob and tell me which one you feel has less line flickering/aliasing?

[Mug Funky]

@ chainmax: sorry, at work atm. but i can say tdeint + blendbob will be faster and probably better if your content is telecined, and especially if it's animated.

@ Uanime5: what colourspace are you working in? if you're in yv12 (DVD source), there shouldn't be a problem. but i haven't tried blendbob on yuy2 (SDI, some DV, or TV capture) yet. it's possible something wacky is going on there. post your script pls :)

[Chainmax]

I'm trying your suggestion and keep getting an "access violation" error pointing to the BlendBob() line. Any idea how to solve that? Also, why using Decimate instead of TDecimate?