Problem with Virtualdub creating Progressive files
 

Hi I am trying to compress (tested all options) an 8 bit interlaced video capture (Matrox) in Virtualdub 1.10.4 using UT Video codec 16.0.2 (422 601BT) but the output file properties when checked in media encoder show it as progressive. Same issue occurs even if i use direct stream copy and check the output file properties. (Mediainfo & Gspot do not show in the properties whether progressive or Interlaced including field order)

Could anyone please guide me on this. Thanks

AVI files don't have a standard flag for indicating interlacing. It isn't a problem.

Checking or unchecking the interlaced option in Ut Video doesn't affect video quality at all, only compression efficiency.

Really thanks for the quick reply. Was confused because when I convert the same file with UT video in Premiere Pro you can retain the field order upper & the same then reflects also when the converted file is opened in premiere (Maybe premiere creates special metadata for its files). While the Virtualdub converted file shows up as progressive when opened in Premiere.

Does this mean that unless specific filters are applied Virtualdub automatically maintains Interlace or Progressive files on as it is condition?
Further it there anyway to force Virtualdub to retain YUV colorspace (without changing to RGB & back) while performing the lossless compression. I think I had read somewhere that fast recompress or direct stream copy doesn't change colorspace but while compressing if I look at VD log it mentions RGB conversion.

Precisely.

You have to ensure that they are treated appropriately. Certain filters, like a Levels adjustment, don't need to know whether a video is interlaced.

Direct Stream Copy will copy the data as-is, and Fast Recompress won't change the colorspace unless the codec you choose forces that. For example, if I open a YUY2 clip and choose the RGB form of Ut Video as the compressor, the log shows RGB24.

Ah but i am actually trying to force Virtuadub to use UTVideo 601BT uly2 since my original capture is uncompressed matrox 8 bit yuv format which means that both are in same color space. It is precisely in this case that virtualdub log shows it is using rgb space even in fast recompress. Thanks

Edit "If i use full compression with default color depth forced to uyuy on both sides for this particular file the decompress log shows xrgb8888 & compress uyuy."

Conversion to RGB as part of Full Processing Mode is expected, as only "recent" VirtualDub versions support YUV filters. Most filters were written before this was added.

If you're getting RGB decompression even with Fast Recompress, I suppose your Matrox decoder is forcing RGB output. Blackmagic's decoders have this problem. One possible solution is to go into VirtualDub's preferences and enable its built-in YCbCr decoders. I think this is off by default.

Using 1.10.4 app. I have the YCbCr on in preferences. So probably the Matrox decoder maybe the one causing this as you say. Thanks a lot for all the clarifications. Amazing site. :-)))

I may be able to find a workaround with an example file.

Create a 1-frame sample of an original capture file with Video set to Direct Stream Copy and Audio set to No Audio. Optionally, compress it to ZIP, RAR, or 7Zip to cut the filesize in half. Attach it to the forum.

Will prepare a small sample tomorrow or day after & load it, thanks.

Do you think in terms of the conversion process there maybe more than a negligible loss in colorspace conversion? If not then it should be fine because I cant see any difference while playing the compressed version on the monitor & tv ie with the original (sources are very old and used vhs cassette recordings). Avisynth could be an option but I have unfortunately never used it before.

Matrox directly supports premiere cs6 but it would use the same matrox codec so am setting my workflow for capture using matrox a/v tools to ensure minimal chance of capture error. Virtualdub will be used to compress to lossless for archival as well for the next stage filter process before going for final edit in CS6.

Planned Workflow
JVC HR 9600 / Panasonic HS860 to MX02mini capture --- Virtual Dub Lossless (UT Video Codec) ---- Resize (masking without scaling Lordsmurf style) ----- Neat Video cleaning + other filters if needed ------- MSU Deinterlace 2.1 (Interlace to progressive) ------ CS6 for edit in terms of still frame replacements for non-reparable content ------ intermediate matrox mpeg2 I frame output (digital master) ------ final compression Handbrake x264 for HDD & private Vimeo.

Does this sequence look good to you? Hoping to become a premium member one of these days & elicit a response from the legendary Lordsmurf on my workflow if he is still around :-))). Seems like I am starting pretty late with VHS conversions.

I'll second msgohan's request for an unprocessed capture sample. Post-processing workflow will differ in many respects for different videos, especially with VHS. But I did notice a couple of things:

Not loss so much as interpolation errors. It depends on when and how conversions are done. VHS should be captured as lossless YUY2 or other 4:2:2 color matrix.

I think you'll find that with VHS, Avisynth is more than an option. There's a first time for everything.

You can't replace lossy encoding problems by decompressing to lossless later. I'd advise that CS6 is for editing but like all NLE's has severe limitations for VHS restoration and cleanup, as any VHS sample will readily demonstrate.

There's more than one way to handle borders and head switching noise. Resizing is neither needed nor recommended.

I use NeatVideo on occasion myself, but it's very limited and causes severe detail loss. It should not be used as a primary denoiser.

Don't do it, excepr for web posting. Deinterlacing always has a cost. The MSU deinterlacer is terrible. If it must be done, there are far better deintelacers.

What kind of frame repair? There are far better tools than CS6 for that.

A short sample will be worth a thousand guesses. :)

Sorry maybe I didn't put it very clearly. I am using matrox a/v tools to capture to interlaced 8 bit uncompressed sd 422 yuv format. (was mentioned in part of the earlier query) Am avoiding CS6 for capture purpose. From this stage I will use the Virtual Dub to fast recompress using UT Vide 422 codec. This interlaced title goes to the archival.

The next phase involves basically preparing an option for personal Web & HDD hence the deinterlacing decision so that I don't have to look at this part later on. --- I would have thought that with the latest viewing media now being primarily progressive, interlace as such was no longer in consideration other than as archival for future upgrade in output quality if required.

Any particular reason that you feel MSU deinterlacer 2.1 is not the right choice. Did try a couple of them but liked the result from MSU. If you can suggest a good one please do let me know thanks.

I have actually tested the following sequence of filters in virtualdub with the results to my liking.
ccd1.7-----Neat Video ----- (Colorfill, MSU color, Brightness/Contrast -depending on what suits) ----MSU Deinterlacer------Resize with only black borders for masking without scaling.

The UT Video output will then only undergo very simple replacement with still frames in case of major errors in CS6 before compression to my digital master (h.264 high quality - handbrake x264 or mpeg2 I frame though am favouring the prior due to final output format)

The content relates to private vhs copies of discourses (with camera position fixed, no movements or scene changes). Thanks

Yes, I understand how you're capturing, my remarks about lossy codecs was simply a caution. You'd be surprised by how many people render to lossy codecs from an NLE and then re-process the video again. LOL!

You should be able to capture directly to losslessly compressed UT video codec without having to recompress.

Obviously you need to deinterlace for most web postings. There are better deinterlacers: yadif with the assistance of other filters, tdeint, and the winner being QTGMC. It would be easy to demonstrate given an unprocessed sample. For final delivery formats to h.264 I'd leave the video interlaced. Deinterlacing is usually done when certain denoisers or repair filters require it, after which the results are reinterlaced.

Avisynth can also replace damaged frames, usually using motion interpolation with a list of specific frames when needed, and can repair frame hopping and similar problems automatically without the trouble of making RGB images. However, the specific type of repair wasn't described. I can think of far easier and faster ways of doing it.

All up you, naturally. Use whatever suits your purpose.

Matrox MXO2 doesn't seem to allow capture via virtualdub hence the initial query of compression colorspace problem during lossless codec encoding. Also wont be able to recapture (too many cassettes each ranging from 2-3hrs) hence the decision to go with matrox a/v tools to ensure zero i/o errors.

I was actually in two minds about the stage of deinterlace but given the progressive nature of todays media display had decided to do it in the initial phase. But as you have suggested will leave the digital master in interlaced condition. Had come across QTGMC the other day in one of the forums, will definitely give it a try.

Also tried TBC 8710 for the best possible capture scenario but it replaces lost signal with last frames from buffer & that causes the video to look unnaturally jerky in the effected place. MX02 seems to hold very well the signal by itself & have set the JVC to edit more with or without DNR depending on the tape output.

The various suggestions on this site have already improved the quality of output so much that was just being lazy about going into avisynth. Plus i find virtuadub extremely capable and the preview facility eases the whole process. Will have a look at avisynth forums for the suggested solutions. Any particular site I should look at?

Thanks for everything.