Any VirtualDub Pros in Da' House?

Movick wrote on 6/19/2010, 2:52 PM
I’m trying to resize some interlaced DV footage in VirtualDub; when performing this process within Vegas, the resizing degrades the footage discernably. My intention is to reduce the 16:9 60i DV footage to 606X404 so that I may overlay a graphical border on top of it for the full frame presentation – approximate 16% reduction. The issue I’m having in VirD is that my processed footage levels are higher than the original. When I drop this VirD processed footage into Vegas I am unable to apply a proper secondary color correction mask to it. I posted to the VirD forums and the developer explained the issue thusly:

“There are several issues involved here.

First, codecs are annoyingly wishy-washy about levels, and programmers have unfortunately often imposed their own beliefs about best practices over any sort of consistency in levels between programs. The most common problem is a mismatch in levels between 16-235 and 0-255 in luma (Y) levels when using YCbCr; less common is a similar problem in RGB space. VirtualDub follows the general convention in Windows of Rec. 601 for YCbCr (16-235) and 0-255 for RGB, and you will get contrast problems when this is not followed by codecs. This is again compounded again by the same kinds of problems in video display drivers. The usual way to diagnose these problems is by switching between 24-bit RGB and YUY2/UYVY in Video > Color Depth until the problem has been identified. If you are working with DV, you can force the internal DV decoder in VirtualDub in Options > Preferences > AVI to avoid at least one source of problems. I've been thinking of adding some helpers (vectorscope, colorbars) to help diagnose this, but haven't gotten around to it. (A vectorscope filter does exist, but it hasn't been updated to support the latest feature set.)”

I’m in a bit of a fix time-wise; the somewhat terse (though much appreciated) explanation the developer offered leaves me with some questions. It seems to take days to get viable replies on that forum and I’m hoping one of the resident VirD experts might be willing to offer me some VirtualDub for Dummies info.

I followed the prescribed steps to force the internal DV decoder. Next I switched the 24-bit RGB color depth to YUY2/UYVY (inout/output respectively) as per the developer’s suggestion.

First: should I match the input and output color depths, e.g. YUY2/ YUY2 or do I set the input to YUY2 and the output to UYVY???

I rendered an uncompressed .avi from VirD with input/output color depth set: YUY2/UYVY. The preview within VirD shows the “after” preview as remaining brighter than the original in the side-by side regardless of what color depth settings I’ve changed. In Vegas, the footage still looks a tad brighter than that which was not processed in VirD; I was however able to create a decent SCC mask with this VirD processed clip.

Any suggestions?

My second set of questions relates to the resizing in VirD. First and foremost, is VirD the best method to resize interlaced footage? And if so, does anyone know the optimal settings within VirD to do so?

Here’s the rest of VirtualDub developer’s comments regarding resizing interlaced DV footage:

“Second, as for the interlaced resize, you can do that in VirtualDub, but it's tricky. A straight resize alone won't give you good results; I recommend trying the following set of filters as a start:

deinterlace (mode: yadif, double)
resize
interlace (frames)

You will need to set the field order correctly on the deinterlace and interlace filters.”

Does the above suggestion make sense? As well, am I better off resizing the footage to the 606X404 within a 720X480 frame and pinning it at the top left corner or am I OK floating the resized footage in the Vegas timeline?

Thanks in advance!!

Mov

Comments

winrockpost wrote on 6/19/2010, 3:09 PM
No help in virtual dub.....but I would do this in Vegas and not expect a serious hit . How are you doing the resizing in vegas ?
Movick wrote on 6/19/2010, 3:20 PM
Event-Pan-Crop.
John_Cline wrote on 6/19/2010, 4:11 PM
Why don't you just do this in Vegas? Make sure you have a "deinterlace method" selected in the project properties. I always use "interpolate." When rendering use the "best" setting. Since you're reducing in size, just use Track Motion to do the resize.
Movick wrote on 6/19/2010, 4:27 PM
Why you ask?? Because the footage doesn't look as good as it does un-resized when done in Vegas. I can see a difference in the VirD footage comparatively, e.g. vertical lines exhibit less artifacting. Believe me, I would much rather hit the ‘Easy button” on this, but the results are not too swell.

See what you think.

http://rapidshare.com/files/400459881/dub_up_blk_full2.mpg.html
John_Cline wrote on 6/19/2010, 5:51 PM
When set to interpolate, Vegas essentially does the same thing as Virtual Dub when resizing. (deinterlace (mode: yadif, double), resize, interlace (frames)) Vegas "unfolds" the frame into individual fields at double the frame rate, rescales the video and then "refolds" (reinterlaces) the fields back into interlaced frames at the new size. The rendering quality must be set to "Best" so that Vegas uses the Bicubic rescaling algorithm. Neither Vegas nor Virtual Dub is actually deinterlacing the video so there shouldn't be any loss of quality other than the loss associated with the unavoidable reduced spatial resolution because it is smaller and the image is represented with fewer pixels.

I didn't think that the MPG you posted looked too bad.
Movick wrote on 6/20/2010, 2:20 PM
John,

Thanks for all the info.

I use “BEST” to render without exception and quite honestly my eyes can’t detect a modicum of difference between “interpolation” and “blend” on PC or after authoring to DVD and playing back on a TV screen. In fact, I created a test disc which included identical segments rendered both ways (blend/interpolation); each appears indistinguishable on TV playback.

I’m still disappointed in the results I’m getting using Vegas to resize the clip – keeping in mind we’re only talking about a sub-16% reduction; the relative degradation IMHO is unacceptable. I went ahead and rendered a clip resized in VirD; vertical lines exhibited less artifacting, but the image looks softer. VirD in experienced hands with the proper tweaking and settings MIGHT pull off a better overall image; unfortunately said hands don’t belong to me and I just don’t have the time to fiddle with the program and its infinite variables.

Between the resizing degradation, the broadcast color clamp washing-out the vibrancy of my formerly crisp, “Technicolor-esque” footage and the mpeg-2 compression, I’ve lost much of the look for which I was going. Coming from a PC designated video background where WYSIWYG, all of the constraints of broadcast-compliant video leave me rather disenchanted to put it politely.

BTW…will using “Track-motion” in lieu of “Event/Pan/Crop” to resize make a difference in image quality? My reasoning for using E/P/C entails the fact that I resize each clip individually as minimally as possible; the 606X404 (approx. 16% reduction) is the MAX I will need to go. A number of clips require less reduction; some are useable at 100% or close to it. As well, a good portion of the video was shot shoulder-mounted; I may need to rotate an occasional clip +/- 1 degree. To use “Track-motion” for this task would require me to create separate video tracks for clips of varying dimensions. Unless there is a benefit quality-wise to using “TM” over “EPC” - I think it best to continue using “EPC” for the resize. If however my image quality will improve, I'll surely switch up to "Track Motion!!"

Thanks again John.

Mov
musicvid10 wrote on 6/20/2010, 3:03 PM
My intention is to reduce the 16:9 60i DV footage to 606X404 so that I may overlay a graphical border on top of it for the full frame presentation – approximate 16% reduction.

That's a 30% reduction in pixels and area, not 16%.
Movick wrote on 6/20/2010, 3:34 PM
30%?????

720 X.16=115.2 | 720-115.2 =604.8. i'm at 606 pixels..

Where on oith did you come up with 30%????? The 16:9 footage represents 873 pixels wide ONLY at 1:1; 16:9 aspect is 1.2121 @720 pixels; the 16% reduction is correct not 30%.

No matter how many pixels at 1:1 aspect - widescreen footage is it is limited to 720 .

Mov
fausseplanete wrote on 6/21/2010, 11:08 AM
AViSynth TDeint (mode 1) = each interlace-field becomes a full frame.