Help with AVI Capture

[rcubed]

Hi,
I started a thread under MPEG2 Encoding, but I think the subject is better discussed here (see http://forum.doom9.org/showthread.php?t=137375
for the details there. I closed the thread over there and moved the request for help here. I don't wish to be cross posting.

The video is encoded in some weird way that has me puzzled. I've had some success with FieldDeInterlace but still see some combing on movement and slightly jerky panning.

Does anyone have some additional suggestions on what to try, I'm getting better with AVISynth, but would call myself challenged to say the least.

I did try Mrestore but am unsure of what parameters to use and what values. The end target is NTSC DVD. I usually encode with HC022.

What I tried was
Tdient(mode=1)
Mrestore(mthresh=0.5,dlcip=last.crop(16,16,-14,12),rate=23.963).

The show is from England so I am not sure what kind of conversion has been done for the US NTSC. As I mentioned in the previous post, apparently something has changed from last year's episodes.

The link to a sample of the source is http://www.mediafire.com/?euzczzhgttg

The captures are with an AVDC-100 (Canopus) so the files are rather large. If additional samples would be of help I can upload them.

Thanks in advance for any help,:)

rcubed

[yup]

rcubed!
Your source NTSC interlaced and Your target NTSC DVD why You need deinterlacing. You plane see this DVD on TV or PC? If on TV You no need deinterlacing. I ordinary make interlaced DVD from interlaced and telecined source, as source contain noise very difficult restore progressive source. For denoising MVDegrainx (x=1,2,3) give better result than FFT3DFilter.
yup.

[rcubed]

rcubed!
Your source NTSC interlaced and Your target NTSC DVD why You need deinterlacing. You plane see this DVD on TV or PC? If on TV You no need deinterlacing. I ordinary make interlaced DVD from interlaced and telecined source, as source contain noise very difficult restore progressive source. For denoising MVDegrainx (x=1,2,3) give better result than FFT3DFilter.
yup.

Yup,
Thanks, I'll give MVDegrainx a try.

If one encodes the video without some sort of filtering, the objects which are moving show what I would call rippling or wavy. Very noticeable on TV. Deinterlacing seemed to fix the problem. If one does a SeparateFields and looks at the separate fields there is blending (combing) present in a single field.

I was wondering if there were any other filters which might be used to fix this problem. In previous seasons, I would simply encode the video with only noise reduction and there was no noticeable rippling. On other captures I've only had to handle telecined input when the material has been telecinded. One of the other series I have captured was apparently done with film. There is something "strange" about this source, and it appears to be unique to how they are processing the video for this series.

rcubed

[Blue_MiSfit]

If you have combed fields you're pretty much SOL :)

You can try vinverse() to clean up residual combing, but good luck..

~MiSfit

[rcubed]

Blue_MiSfit,
Thanks for the suggestion, I'll give it a try. I've never seen combed fields before. Makes me wonder what they did to the source to get that.

rcubed

[Blue_MiSfit]

Hehe.. I see it frequently.

The most common culprit is resizing interlaced content like it's progressive - which destroys the interlacing! :)

~MiSfit

[45tripp]

what's with all the speculation?

it's not interlaced, or oddly telecined,
and there are no combed fields.

it's from England yeah,
so 25 fps probable,
an so.

use a simple
tdeint(1)
mrestore()

for going to 23.976,
in case you feel like trying it, or need it in future
tdeint(1)
mrestore(numr=40,denm=100)

mthresh is not a paramater of the mrestore i have...
it's mthr and it takes ints


for something more complex try maybe:
tdeint(1)
a=last.dfttest(sigma=2).crop(4,8,-4,-8)
mrestore(mthr=-8,chroma=true,dclip=a)

and,
pick a scene with some more motion in it..
to test.

gl

[rcubed]

Immerson,
Sorry I used the wrong term :stupid:, as you pointed out in the other forum s/b field blending.

What is the latest version of Mrestore? The 1st one I had found was v1.6a 11/22/2006. Since then I have found a version 2.3C dated 12/8/2007. The version 1.6a did have parameters mthresh and bthresh. I notice the newer one does not. Is the 2.3c the version to which you refer? Both were contained in a zip file R_pack by mOmonster. If not the most recent do you have a link?

When I use the defaults the result is 25fps, which is non compliant for an NTSC DVD, hence the 23.976 and using 3:2 pull down in HCEnc. Is there a better way to treat it?

Thanks for the clarification and help. I apologize for using the wrong term.

rcubed

[45tripp]

If not the most recent do you have a link?


it's the right one.


When I use the defaults the result is 25fps, which is non compliant for an NTSC DVD, hence the 23.976 and using 3:2 pull down in HCEnc. Is there a better way to treat it?


yes,
encode as 720x480 @ 25fps
use dgpulldown to pulldown 25->29.97

[rcubed]

Immersion, (sorry about the typo last post)

I tried the

tdeint(1)
mrestore()

encoded as progressive and then used dgpulldown to pulldown 25->29.97.

The final output looks very good on TV, nice smooth pans and object motion. :)

After reviewing what little data I had laying around for last season, it looks like I was using ProcoderExpress 1.2 to do the encodes on this particular series. So I was probably wrong about things changing in the source this season. I tried the sample on PCExpress and it didn't have any of the nasty artifacts I had been getting prior to your help. At some point I had switched over to using AviSynth and HCEnc apparently after the series was done. The default settings in Procoder seem to be able to handle the source pretty well (not that there are a lot of options with the Express version).

I'm currently processing the 1st two episodes with mrestore. The only down side is it's rather slow even on my AMD 5600+ X2. I'll just let it run overnight. :) I'm running two instances of HCEnc and the two processors are pretty well pegged at 100% doing about 9.6 fps (avg) on each. (In the winter it helps heat my computer room) Quality over speed in my book!

Again I appreciate all the help you have supplied. Each time I learn a little more. Thanks again for taking the time and effort. :thanks:

rcubed