Capturing PAL sources with capture card... Best way???

[Dat]

Hi, i got my capture card quite recently and am just starting to get to grips with it properly.

Im trying to find the best method of capturing a PAL (interlaced) source in order to convert to SVCD, i have read lots of guides (most only give details for NTSC sources) but i still wonder what other methods people are using because im sure i can get better quality encodes.

Im unsure what is the best framesize to capture at? if to use any filters in VDub? What capture codec?

And as i said i have read loadsa guides, i just wanna know from people personal experience so please dont tell me to read some more guides etc.

Thanks in advance,

Dat :D

[sherpya]

I'm using an Osprey capture card (osprey are cheap cards because they have no hardware encoding/decoding engine), so I capture directly in YUV2 with VirtualDub. I'm using huffyuv codec but It's not always usable (You will need a lot of space).
Over the capture avi I use TomsMoComp avisynth filter. You can use other filters depending on quality of the source (like TemporalSmoother, SmootherHiq). The make your own avs file and Encode it with VirtualDub.
For the resolution you can use Full Pal, if the movie is short (24 mins are ~ 12 gb of avi) and then crop & resize to a lower resolution. This applies for best quality. You can also choise the resolution directly on the source format card setting using a lower res. However your card can come with hardware MJPEG encoding/decoding capabilities (most likely) and you can use MJPEG in place of huffyuv.
Quality is not the same, but is acceptable.

[Dat]

Originally posted by sherpya
I'm using an Osprey capture card (osprey are cheap cards because they have no hardware encoding/decoding engine), so I capture directly in YUV2 with VirtualDub. I'm using huffyuv codec but It's not always usable (You will need a lot of space).
Over the capture avi I use TomsMoComp avisynth filter. You can use other filters depending on quality of the source (like TemporalSmoother, SmootherHiq). The make your own avs file and Encode it with VirtualDub.
For the resolution you can use Full Pal, if the movie is short (24 mins are ~ 12 gb of avi) and then crop & resize to a lower resolution. This applies for best quality. You can also choise the resolution directly on the source format card setting using a lower res. However your card can come with hardware MJPEG encoding/decoding capabilities (most likely) and you can use MJPEG in place of huffyuv.
Quality is not the same, but is acceptable.

What resolution do you actually capture at then? and do you recommend i try to capture at SVCD resolution or more then resize to SVCD?

Thanks :)

[bb]

For best results capture at full PAL res. Then filter, then resize down to SVCD res and encode.

If you ask me, you can capture at SVCD resolution directly, then filter and encode. The result will be almost as good as the above method, but much less time and HDD space consuming.

bb

[Didée]

( @ bb )

The question is: If one captures in lower resolution than MAX - how is the resizing handled by the capture card?
I guess this is different from card to card, but 1) I have only experience with 2 different ones, and 2) in neither of both documentations I found any hints about this resizing aspect ...

Personally, I don´t like anything other than 720x576 with Huffyuv & PCM audio (though I do mpeg-2 capturing occasionally, HD space is rare sometimes...)

"If you want something to be done right, do it yourself!"

[bb]

First I'd like to make clear we are talking about horizontal resizing only. Vertical resolution should always be 576 for PAL or 480 for NTSC.

If your TV card provides higher horizontal resolution, then I guess it will do this by changing the sampling rate of the signal. Thus you should get a more accurate scan. The noise will be scanned more accurately, too, but as you need to downsize to 480, the noise will be decreased.

If you capture at a lower sampling rate, i.e. lower resolution like 480, then the noise reduction effect you'd get by downsizing will be gone. So there will be more noise this way - if you really recognize that is another question.

The spatial and temporal filtering will have a bigger effect IMHO.

bb

[theReal]

For best quality, capture at full resolution in YUY2 (Huffyuv or MJPEG), resize with a filter that makes the content compressible (like Precise Bilinear in VDub) and add some noise filters. I can recommend cnr2 for avisynth (it's "the same" as chroma noise reduction for VDub), temporalsmoother(VDub and avisynth) and smoothhiq2(avisynth).

Also a good idea is to filter chroma and luma seperately, here's how to do it in avisynth (based on an idea/script by Acaila)

#############################
Source=avisource(G:\movie.avi).BicubicResize(640,272,0,0.5)

#apply temporal smoother on the source clip
Temporal=Source.TemporalSmoother(2,1)

#apply spatial smoother (or cnr2) on the source clip
#then discard the luma and use luma from the above clip "Temporal"
Spatial=SmoothHiQ(Temporal,3,25,10,192,3).MergeLuma(Temporal)

#output the final clip
Return Spatial
##############################

This way you can use stronger filter settings, especially on the chroma, your video will be more compressible but it will still not look extremely filtered.