VideoHelp Forum




+ Reply to Thread
Results 1 to 11 of 11
  1. Member
    Join Date
    Mar 2002
    Location
    Croatia
    Search Comp PM
    Hi!
    Is it possible to capture from VHS and burn as VCD but without loosing one field of picture? I know that it's possible with SVCD because it is captured with 480x576 (PAL, both fields, interlaced). I ask this question because on VCD I get picture like from the movie camera because I loose one field. The picture is not fluid on fast movements of camera.
    Thanks...
    Quote Quote  
  2. Member flaninacupboard's Avatar
    Join Date
    Aug 2001
    Location
    Northants, England
    Search Comp PM
    Something you can try doing is capturing at 352X576 using the TMPGenc deinterlace filter "Double" and resizing to 352X288, this will give you one frame averaged from the two fields.
    Quote Quote  
  3. MPEG1 does not support interlaced video; however, you can choose to capture both fields (352x480) and then using VirtualDub, deinterlace (use the smart deinterlace filter) and then resize using the smart resize filter. This will combine both fields and give you twice the data to work with as well as a more fluid looking video. You can chose to blend (fast motion frames may look kinda blurred) or lose just one field on only the area of the frame that has the "combing" artifacts. I make all my VCDs this way now and they look great.

    Brian
    Quote Quote  
  4. Member
    Join Date
    May 2002
    Location
    australia
    Search Comp PM
    if you capture at the higher resolution (480) and use tmpgeng to encode to mpeg1 using the default vcd template and no filters, tmpgeng appears to handle the interlacing automatically ... but for best quality, are you saying I should manually specify a deinterlace filter either via virtualdub or tmpgeng or is tmpgend doing this ok with the vcd template? whats your recommendation ... thanks.
    Quote Quote  
  5. Member flaninacupboard's Avatar
    Join Date
    Aug 2001
    Location
    Northants, England
    Search Comp PM
    if you're captruing at X480 i guess you're in NTSC land (i'm in glory-PAL) so it depends on whether you capturing something that was recorded at 29.97FPS or something transcoded to 29.97FPS. Films are recorded at 24FPS and transcoded to 29.97 for TV transmission or on laserdisc or vhs. If its a film you're capping you need to use a 3:2 pulldown to avoid the end result looking jerky. I never need to do this because i always use PAL, so know nothing about it. if you're capping things like TV series or made for TV movies, they're more likely to have been shot on video so will inherently be 29.97FPS.
    I've always found that when reducing from 480(576) to 240(288) hence chucking an entire field, that its better to use the deinterlace filter double then resize, and while it may look "smoother" on the monitor, it will look better when viewed from a normal TV type distance than just ditching a field.
    Quote Quote  
  6. I haven't really compared the TMPEGenc deinterlace filter with VDub's smart deinterlace filter. What is important is how TMPEG deinterlaces... do they blend fields or simply toss one of the fields? If it is tossed, then you might as well just capture one field at 352x240.

    I personally find the filters in VDub give better final quality clips and seem to be faster.
    Quote Quote  
  7. Is your source a DV camera? There's really no satisfactory conversion from DV to VCD. Deinterlacing will not help--and it can make things worse by introducing the illusion of wierd double-images.

    The problem is not intuitive, and the explanation is a bit esoteric.

    The eye detects motion when there's evidence of "apparent motion". (Yes, that's the technical term for it!) Apparent motion can include spatial cues, like blurring, or temporal cues like frequent visual changes.

    Because it stores a non-interlaced picture, VCD eliminates the temporal aspect of apparent motion. Every image in a VCD actually gets flashed onto the TV screen twice, so VCD depends on spatial cues (blur) to communicate motion to the eye.

    But DV cameras eliminate the spacial aspect of apparent motion; every DV field is a perfectly blur-free picture. Every interlaced frame contains two of these, one followed by the other, so DV depends on temporal cues (movement between fields) to communicate motion to the eye.

    When you record DV to VCD, you lose both apparent-motion cues. The result is truly bizarre: it looks like the picture is stopping and starting thirty times a second. Motion looks extremely rough, and the eye becomes strained.

    The best recourse for video recorded from a DV camera is to use SVCD, DVD, or tape. VCD may be marginally acceptable for material that doesn't have much motion.
    Quote Quote  
  8. Member flaninacupboard's Avatar
    Join Date
    Aug 2001
    Location
    Northants, England
    Search Comp PM
    sorry tacosalad, but you're just plain wrong!
    If you cap the DV to your hard drive, youll have 720X480 interlaced, and both fields are different. now, if you load this clip into TMPGEnc, and use the deinterlace filter "Double" it blends the two fields together, which generates the blurring required to convince your eye it sees smooth motion. in low/no motion the deinterlacing is not apparent, as there isn't usually a huge difference between fields, and when there is high motion the blur generated by the field merging is perfectly convincing! try it!
    Quote Quote  
  9. Heck, I wouldn't mind being "just plain wrong". I offered your advice to my brother, my brother-in-law, and fellows on this board several years ago. It didn't work, and one fellow finally sent me a clip of just a few frames that revealed why: individual DV fields don't have any blur, so blending them just yields a double-exposure effect.

    After this loss of face, I looked into a few of my textbooks. My comments above were a layman's translations of discussions by Adelson and Bergen (E.H. Adelson and J.R. Bergen "Spatiotemporal models and the perception of motion" Feb 1985), and Watson and Ahumada (A.B. Watson and A. J. Ahumada, Jr. "Model of human visual motion sensing", Feb 1985). If you have access to a well-stocked academic library, there's an excellent illustration in the text book (not the standard itself) "MPEG Video Compression Standard" by Joan L. Mitchell, William B. Pennebaker, Chad E. Fogg, and Didier J. LeGall, pages 75-79. In particular, figure 4.15b illustrates the effects of temporal quantization (as produced by VCD) and in figure 4.15c demonstrates what would happen if it were combined with spatial quantization (as from a DV camera). If I got it wrong, either my sources are wrong or I translated it badly into lay terms. (Admittedly, both are possible.)

    The problem actually occurs because of the high-speed shutter in a DV camera, so you could solve it if you could disable the shutter. Is this possible? All the DV cameras I've seen don't allow the shutter to be turned off, but I suppose there could be one. I didn't mention that.

    Another solution would be a filter that re-creates natural motion blur, but I haven't seen one. Not for free, anyway. TMPEG's deinterlace filter uses fixed-coefficient bilinear interpolation, which is obviously inadequate. (It also offers non-interpolating modes, but those don't help.) Come to think of it, that sounds like an interesting project! I've written a number of codecs and filters, but never published any since they were all designed for specific needs of my own. (Do you have any use for a temporal-delta signed-gamma-distribution compression codec? Heck! Who else would??)

    But I digress. By all means, try deinterlacing or anything else, and let me know if you find a way that works. It'll make me a hero with my brother-in-law!

    -tacosalad

    PS: Sorry if my patience sounds thin. It isn't you. It's unbelievable how much modem trouble I've had while trying to post this message!
    Quote Quote  
  10. DV cameras (and indeed all video cameras) run "shutter" speeds (actually refresh rates) of 50Hz (50 times a second) - or 60Hz if you are in America & Japan. You can set higher or lower speeds manually but this will affect the picture quality. I have never seen any problems with importing or converting standard DV footage. I don't see how this would affect anything to do with VCD conversions.
    Quote Quote  
  11. There is a temporal blurring filter in virtualdub -- and it is quite customisable too...

    I've used it once before to create a more realistic "motion blur" effect for a computer generated fly-through clip (into clip on the Demo VCD): http://www.vcdimager.org/discs.phtml

    Regards.
    Michael Tam
    w: Morsels of Evidence
    Quote Quote  



Similar Threads

Visit our sponsor! Try DVDFab and backup Blu-rays!