-------------------------
MPEG2AVI v0.16B3 10/15/99
-------------------------

(Since v0.15B3, there have been no changes to the filtering operations.
 The fast-filter is now implemented with MMX instructions, making it
 much faster than HiQ mode.)

These notes are a bit terse.  They will be re-organized for the
next release.

        MPEG2AVI fast filter vs. HiQ filter
        ----------------------------------------------
              v0.15B1            |        v0.14
                                 |
        half-res   |    full-res | half-res   full-res
          '-2'     |             |   '-2'
        -------------------------|--------------------
        *HiQ  '-q0'  *HiQ  '-q0' |  *fast       *HiQ
         fast '-q1'   fast '-q1' |
        
           * = default
        
        
        MPEG2AVI v0.15 offers two filters for its normal (full-resolution) and
        special half-res mode.
        
        ---------------------|--------------------------------
        fast-filer HALF-RES  |  arithmetic average on 4 Y-pixels
          '-q1'       '-2'   |  no filtering performed on U/V pixels
                             |
                   FULL-RES  |  pixel replication on U/V pixels
                             |  (1 U/V pixel => 4 identical U/V pixels)
                             |  no filtering on Y plane
        ---------------------|--------------------------------
         HiQ-filer HALF-RES  |  low-pass FIR filter on Y-plane
          '-q0'       '-2'   |  no filtering on U/V plane
                             |
                   FULL-RES  |  upscaling interpolative FIR filter on U/V plane
                             |  no filtering on Y plane

---------------------
Usage recommendations : "fast filter" vs "HiQuality"
---------------------

        "RGB" vs "YUV"
        All PC displays store color-information in the "RGB" format.  A pixel
        is divided into three additive color components, red green and blue.

        Human visual perception operates differently.  First of all, human
        eyes are more sensitive to brightness (luminance), than to color
        information.

        Therefore, a color-format which partitions pixels by brightness/color
        is better for video-compression.  Nearly all VideoForWindows codecs
        discard 1/2 the color (U/V) data during input.  When encoding with
        these codecs, postfiltering the decoded MPEG pictures' U/V plane is
        a waste of time

        Thus, MPEG2AVI's "fast-filter" will produce output faster, while
        maintaining comparable video quality (depends on the codec.)
        
        MPEG2AVI's HiQ-filter should be used if reserved for special situations, like
        uncompressed AVI mode, where maximum quality is desired.

-------------------------
MPEG2AVI and downsampling
-------------------------

        As an exception to the above, the use of downsampling ( -3X or -3Y )
        necessitates the HiQ-filter.

        MPEG2AVI v0.15 implements a simplistic downsampler: an arithmetic
        pixel averager.  Output pixel values are calculated as weighted
        means of input-pixels, with a weighting policy controlled by
        Bressenham's line-algorithm.  The mean-weighted outut
        simulates low-pass filtering, but the filtering is less
        effective against bitmaps consisting of replicated pixel blocks.

        Pixel-averaging is fast, and smooth for scaling factors between 
        0.5 and 1.0.  There is only 1 downsampling algorithm : the 
        filter-quality option ('-q0', '-q1') does not apply to downsampling,
        only pre-filtering.
        
        All downsampling algorithms diminish image detail.  Simultaneous
        X & Y downsampling should be avoided if possible.