Interlacing Problem: Tsunami vs. CCE SP

[Karlson]

Dear all,

I tried to encode MPEG-2 with both, Tsunami and CCS SP. The material was captured with Matrox Marvel 400 TV. It consists of several files, I therefore used AviSynth to concatenate them and also to create 480x576 PAL SVCD size.

Method 1 works well: Feed it directly to Tsunami and encode. I de-interlace with Tsunami's built-in function, selecting odd. The result looks quite impressive. No problems.

Method 2 has a problem: I use VirtualDub to convert it to the correcvt size, as CCE SP isn't able to do so. I recompress using either HuffYuv of PICVideo codec with quality setting 20. By doing so, I also noticed the interlacing artefacts. The result will be that in scenes with a lot of motion, there are some ghost structures.

I tried to de-interlace with seveal VD filters, including smart-deinterlace. I tried different settings of CCE SP: but in the end, I think this is not the correct way. I thought the card is producing interlaced video, odd.

Shouldn't there be a quick and simple way to de-interlace without motion estimation etc. like in smart-deinterlace?

Thanks for every hint.

[Sciascia]

Have you tried Frameserving from V-Dub to CCE SP?

It is pretty simple if you can do AVI Synth than this should be cake. Instead of recompressing choose uncompressed as your output codec and your audio as direct stream copy. Apply all of your video filters; ie. resize, deinterlace, 2D cleaner, whatever and start frameserving it to CCE. Make sure that video is set to full processing mode. It is still faster than TMPGEnc (as far as I remember) and the qualty is great. It works especially well with the MJPEG codec of the Marvel. And of course you can use the Append segment to join several clips together.

[Karlson]

Thanks for the tip. I'll try that. But on the other hand: you spoke about "deinterlace". That's exactly my problem.

You seem to have experience with that scenario. So what settings did you use?

Thanks.

[Sciascia]

It's been a while, but here are the settings as far as I can remember:

1. Resize to 480*480 using precise bicubic

2. Smart deinterlace (low setting using the preview option till it's clean)

3. A little 2D cleaner

4. Uncompressed, full processing mode for video, direct stream copy for audio

5. Frameserve to CCE SP

Hope this helps, use the preview window to see what it is going to look like going to the encoder. The fewer processing that is going on the better the results are going to be.

[Karlson]

Thanks. I'll give it a try.

It's Baby Video. So it may take a while before I have time. As you are German, you might have read in c't already in a reader's comment that he said: only unemployed and retired are realy able to do video editing

[mikiem]

Hi Karlson

You didn't mention if you selected the box to create two fields within picvideo. If so, you're possibly interlacing your deinterlaced video.

mikie

[Karlson]

Thanks for all the help. I wasn't able to try out all of them so far. A short summary: in the beginning I thought that the source would be captured with 576 lines that are afterwards split into two (odd and even numbered lines). I thought Tsunami would simply re-arange them into one 576 line image.

Donald Graft (Mr. smart-deinterlace) explained to me that Tsunami will only select odd lines and double them, thus reducing the horizontal resolution. And in fact I was able to see that in the picture. For such a reduction, the result was pretty good.

So I now used the smart-deinterlace and produced the video with CCE SP and the result was (clearly) much better, because now I use 576 lines vertical resolution.

But in any case, I would have expected that there is a switch to say to CCE SP: source is interlaced, especially because I thought the output is also interlaced. So there should be no real problem. I am still a bit confused and will try other tips.

Thanks again.