Announcement

Collapse
No announcement yet.

Opinion on NTSC and YUY2/Huffyuv?

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

  • Opinion on NTSC and YUY2/Huffyuv?

    Hello,

    I have been really experimenting with the YUY2/Huffyuv setup on the Marvel G400. From my take, the quality of resulting files is very good. In previous projects, I have used Matrox MJPEG.

    I am having a hard time trying to distinguish quality differences from using YUY2/Huffyuv and Matrox MJPEG (highest quality setting). Sometimes, I think MJPEG provides better quality, then again sometimes I see artifacts.

    There are a lot of people on this board that have been doing this for awhile (thank you). My goals are to produce videos delivered on Hi8, VCD, and SVCD. With the issues of PAL aside (I am using NTSC), what is your opinion on which codec is better? How would you really tweak out the Marvel G400 to give you high quality without crushing the hard drives? Any opinions are greatly appreciated.

    FYI, here is my system:
    FIC AZ11 Mainboard
    AMD Athlon 1.1GHz
    128MB RAM
    2-20GB HDD
    Matrox Marvel G400
    SBLive
    Creative Dxr3 Decoder
    16x Panasonic DVD
    Mat. CDRW
    OS: Win98

    Thank you

  • #2
    YUY2 is the better source because it has a greater dynamic range (using photographer-speak) due to it being an uncompressed format with a 4:2:2 colorspace. Huffman encoding (HuffYUV), while making the file smaller, does not quantize (reduce the number of) the colors in the image so the dynamic range remains intact.

    MJPeg, DV and all other DCT based codecs quantize the color information when they are initially captured, which reduces the dynamic range. MPEG compression then quantizes the data again before encoding the data. This can be murder on the dynamic range in blurred backgrounds, shadows and highlight areas of the image.

    This re-quantization causes a problem because the original DCT based codec and MPEG very likely will be using different quantization matrices to do the job.

    One of the side effects of this are the artifacts you often see in MPEG's that are created from DCT compressed sources. The greater the degree of quantization the greater the chance of block artifacts in the affected areas noted above.

    DV is triple-damned because it starts out with a reduced colorspace (4:1:1), which gives it a reduced dynamic range to start with. Encode that using DCT then re-quantize it again with MPEG and...you get the idea.

    Using an encoder like TMPGEnc can help when encoding DV. It has a menu item which will throw it into a 4:4:4 (full color) emulation mode. While it isn't the same thing as having a higher dynamic range to start with it at least takes into account DV's limitations and corrects some of the problems.

    Editing on a device like the RT-2000 also helps as it converts DV to a 4:4:4:4 RGBA colorspace internally for compositing its effects. If one then frameserves (by way of AVISynth) the timeline to TMPGEnc for encoding the results can be very good for a DV => MPEG conversion.

    Dr. Mordrid



    [This message has been edited by Dr Mordrid (edited 19 February 2001).]

    Comment


    • #3
      Wanting both analog and MPEG output formats makes it tough to do it all only one time. HuffYUV would be great except it seems very difficult to get the HuffYUV codec to do analog output that you can record. At least the w2k G450 drivers don't play DVDmax on HuffYUV. (PICvideo MJPEG does play my old DMB1 G200 Marbel captures out DVDMax so I can still make tapes from them if necessary).

      If you have the hard disk space You'd get the best quality by capturing and editing in HuffYUV then transcoding (final rendering, make movie, create video file, etc.) to the Marvel's DMB1 MJPEG codec for analog output over the SVHS connector.

      If you think the MPEGs made form MJPEG captures are of sufficient quality then doing everything in DMB1 MJPEG is the the easiest in terms of both time and hard disk space.

      The 2/4GB file limt will be a hassle on win98. I've been using w2k for the past year so I can't offer any advice as to workarounds that won't disrupt your workflow.

      I do everything in DV now, analog captures with HuffYUV isn't really worth the trouble for me since:

      1) my cable company doesn't deliver that great a picture, so a straight analog to DV capture thru my D8 camcorder gives me about the same MPEG results -- perhaps if there were w2k Marvel drivers it'd be different as my cheapo Hauppauge WinTV lacks SVHS input.

      2) Analog captures of my DV camcorder only adds analog noise as the color space is already reduced when recorded to DV tape originally.

      The good news is I can produce a ~35 min DV output file (~7GB, no problemo on w2k with NTFS disk format) and encode it straight to VCD or SVCD with TMPGEnc or send it thru the camcorder and make analog tapes. Despite needing the extra hard disk space this is for me the best compromise of cost vs. my time.

      --wally.

      Comment

      Working...
      X