Announcement

Collapse
No announcement yet.

Could use some Help, RR_G - YUY - MPEG1(VCD)

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Could use some Help, RR_G - YUY - MPEG1(VCD)

    I have been capturing in YUY format with VT 1.52 with no compression. This is done in the 352x288 format. When converting with the Panasonc 2.03 MPEG1 encoder, the resulting output is just plain nasty. Blocks everywhere during the ocean scene. If I up the bit rate (3000+) and make non-standard VCD MPEG1 everything looks good and would be acceptable to output and run in my set-top DVD player and TV.
    ftp://swarm.iai.net/swarmftp/FTP.Dow...st_samples.zip

    _3 is the 5000+ bit rate
    _1 is the VCD standard bit rate

    Ok, here are the two output examples in MPEG1 (if anyone can help with this challenge). I also have the raw YUY output, but not sure that it is necessary for this test. Can ANYONE take the "test_sample_YUY2_352_3.mpg" sample file and make it into a VCD compatible file without all the nasty pixelization or can this just not be done?

    I have seen quite a few professional VCD's that look just great with water scenes, why is it that I am unable to do this?

    Questor


    [This message has been edited by Questor (edited 26 June 2000).]

  • #2
    Ahh...the joys of low datarate MPEG encoding. Such fun.... Now you know why folks are so anxious for SVCD to come into its own.

    Actually with the exception of the Gibbs Effect stuff (squiggles around the perimeter of humans, text & other solid objects in the foreground) the artifacting doesn't really get out of hand until the scene moves to include the sunlit water and waves, does it? This is the clue.

    The fast and random motion in the waves combined with the bright flashes from the sun are the main culprits. Also some, but not all, of the blocking in shaded areas is because of the high level of contrast in the scene.

    Shooting tip: the next time you shoot a scene with water and/or sun in the background try using the backlight exposure control on the cam. Once set it exposes for the foreground shadow areas instead of being fooled by the background into underexposing the scene. It doesn't look like you used it here.

    Problem scenes like this require the so-called shotgun technique. That is, hit it with everything you've got. Even so you may not get great results with most consumer software. This is a difficult scene and the commercial grade encoders are MUCH more sophisticated than anything we get to play with when it comes to compensating for such things.

    That said the first suggestion is to click on the Advanced Settings button and turn up the motion estimation to Full Pel. Higher motion estimation settings are used where fast moving subjects (such as waves, fast cars & rampaging 2 year olds) cause artifacting.

    While you're there try setting the Noise Reduction to Strong, the Video Filter to Adaptive/Strongest and the Color Tone Correction to TV. This should help with the Gibbs Effect and perhalps the high contrast and the bright flashes. Like I said....shotgun.

    The other file is better just because of the higher data rate. This helps offset some of the problems caused by motion and reduces the Gibbs Effect artifacts considerably.

    As far as re-encoding the high bitrate file into a good looking VCD file....dream on. Most times this just introduces a whole other catagory of problems: recompression artifacts. 'tis far better to use higher quality settings and re-encode the original YUY2 file.

    Dr. Mordrid


    [This message has been edited by Dr Mordrid (edited 27 June 2000).]

    Comment


    • #3
      Thanks for the help, this seems to have helped greatly. What would be the result of using these setting for all YUY - MPEG1 projects?

      Comment


      • #4
        As long as you don't mind the longer rendering times I guess you could use those settings all the time.

        The main problem here was the high contrast combined with the fast moving and flashing background the waves & sun presented. Absent that kind of problem you could probably get away with just using the Noise Reduction at lower data rates to reduce the Gibbs artifacts.

        Dr. Mordrid


        [This message has been edited by Dr Mordrid (edited 27 June 2000).]

        Comment

        Working...
        X