vp3-format.txt

mediastreamer2是开源的网络传输媒体流的库

                        discard up predictor
                        pred = (53 * ur.dc) + (75 * l.dc)
                               --------------------------
                                         128

     1   0   0   0    up-left predictor available:
                        pred = ul.dc

     1   0   0   1    up-left, left predictors available:
                        discard up-left predictor
                        pred = l.dc

     1   0   1   0    up-left, up-right predictors available:
                        pred = (ul.dc + ur.dc)
                               ---------------
                                      2

     1   0   1   1    up-left, up-right, left predictors available:
                        discard up-left predictor
                        pred = (53 * ur.dc) + (75 * l.dc)
                               --------------------------
                                         128

     1   1   0   0    up-left, up predictors available:
                        discard up-left
                        pred = u.dc

     1   1   0   1    up-left, up, left predictors available:
                        pred = (-26 * ul.dc + 29 * u.dc + 29 * l.dc)
                               -------------------------------------
                                                 32

     1   1   1   0    up-left, up, up-right predictors available:
                        pred = (3 * ul.dc + 10 * u.dc + 3 * ur.dc)
                               -----------------------------------
                                                16

     1   1   1   1    all 4 predictors available:
                        discard up-right predictor
                        pred = (-26 * ul.dc + 29 * u.dc + 29 * l.dc)
                               -------------------------------------
                                                 32

Note that this final prediction case ([ul u l]) risks outranging. The
difference of the predicted DC is checked against u.dc, l.dc, and ul.dc,
in that order, and if the difference is greater than 128 in any case,
the predictor is assigned as that DC coefficient. In pseudocode:

  if (ABSOLUTE_VALUE(pred - u.dc) > 128)
    pref = u.dc
  else if (ABSOLUTE_VALUE(pred - l.dc) > 128)
    pref = l.dc
  else if (ABSOLUTE_VALUE(pred - ul.dc) > 128)
    pref = ul.dc

The predicted value is, at long last, added to the fragment's decoded DC
coefficient. Finally, the new DC coefficient is saved as the frame
type's auxiliary predictor. For example, if this fragment is coded as a
motion residual from the previous frame, save the fragment's DC
coefficient as the previous frame auxiliary predictor.


[still need to mention precise rounding considerations, a.k.a, the
HIGHTBITDUPPED() macro]



Reconstructing The Frame
------------------------
rough outline:
  - foreach fragment:
    - if motion vector
      - copy motion fragment from appropriate frame into current frame
        (don't forget to account for unrestricted motion vectors)
    - dequantize fragment coefficients
    - run coefficients through inverse DCT
    - if INTRA coded fragment
      - output transformed coefficients
    - else
      - apply transformed residual to motion fragment
    
[not finished]


Theora Specification
--------------------
The Theora project leverages the VP3 codec into a new video coding
system. The algorithm and bitstream format are the same as VP3 with a
few minor differences:

1) The frame orientation is reversed-- VP3 is coded from bottom to top
while Theora video is coded from top to bottom.
[nope-- only true in the first few alpha releases; final Theora spec will
be upside-down, the same as VP3]

2) Variable histograms-- VP3 uses a hardcoded set of histograms for DCT
coefficient coding (described in section "Unpacking The DCT
Coefficients"). Theora packs the histogram information in the header of
the transport format (which is meant to be Ogg, but can probably be
coerced into a variety of other multimedia container formats).

3) Variable quantization-- As with the histograms, Theora codes the
quantization tables and quality thresholds (described in section
"Initializing The Quantization Matrices") into the header.

4) [special VLRLC case for encoding unusually large runs of blocks;
necessary for HD resolutions]

[still need coding format of histogram and quantizer information]


Appendix A: VP31 Quantization Matrices And Scale Factors
--------------------------------------------------------
The following quantization matrices and scale factor tables are hardcoded
into the VP31 coding standard. These tables can vary according to the
setup information transported along with a Theora file.

Base quantization matrix for golden frame Y fragments (note that this 
is the same as JPEG):

    16  11  10  16  24  40  51  61
    12  12  14  19  26  58  60  55
    14  13  16  24  40  57  69  56
    14  17  22  29  51  87  80  62
    18  22  37  58  68 109 103  77
    24  35  55  64  81 104 113  92
    49  64  78  87 103 121 120 101
    72  92  95  98 112 100 103  99


Base quantization matrix for golden frame C fragments (note that this 
is the same as JPEG):
    
    17  18  24  47  99  99  99  99
    18  21  26  66  99  99  99  99
    24  26  56  99  99  99  99  99
    47  66  99  99  99  99  99  99
    99  99  99  99  99  99  99  99
    99  99  99  99  99  99  99  99
    99  99  99  99  99  99  99  99
    99  99  99  99  99  99  99  99


Base quantization matrix for interframe Y and C fragments:

    16  16  16  20  24  28  32  40
    16  16  20  24  28  32  40  48
    16  20  24  28  32  40  48  64
    20  24  28  32  40  48  64  64
    24  28  32  40  48  64  64  64
    28  32  40  48  64  64  64  96
    32  40  48  64  64  64  96 128
    40  48  64  64  64  96 128 128


DC coefficient scale factor table:

   220 200 190 180 170 170 160 160
   150 150 140 140 130 130 120 120
   110 110 100 100  90  90  90  80
    80  80  70  70  70  60  60  60
    60  50  50  50  50  40  40  40
    40  40  30  30  30  30  30  30
    30  20  20  20  20  20  20  20
    20  10  10  10  10  10  10  10


AC coefficient scale factor table:

   500 450 400 370 340 310 285 265
   245 225 210 195 185 180 170 160
   150 145 135 130 125 115 110 107
   100  96  93  89  85  82  75  74
    70  68  64  60  57  56  52  50
    49  45  44  43  40  38  37  35
    33  32  30  29  28  25  24  22
    21  19  18  17  15  13  12  10


Appendix B: Macroblock Coding Mode Alphabets
--------------------------------------------
These are the 6 pre-defined alphabets used to decode macroblock coding
mode information:

Alphabet 1:
  index   coding mode
  -----   -----------
    0     MODE_INTER_LAST_MV
    1     MODE_INTER_PRIOR_LAST
    2     MODE_INTER_PLUS_MV
    3     MODE_INTER_NO_MV
    4     MODE_INTRA   
    5     MODE_USING_GOLDEN,      
    6     MODE_GOLDEN_MV   
    7     MODE_INTER_FOURMV

Alphabet 2:
  index   coding mode
  -----   -----------
    0     MODE_INTER_LAST_MV
    1     MODE_INTER_PRIOR_LAST
    2     MODE_INTER_NO_MV
    3     MODE_INTER_PLUS_MV
    4     MODE_INTRA
    5     MODE_USING_GOLDEN
    6     MODE_GOLDEN_MV
    7     MODE_INTER_FOURMV

Alphabet 3:
  index   coding mode
  -----   -----------
    0     MODE_INTER_LAST_MV
    1     MODE_INTER_PLUS_MV
    2     MODE_INTER_PRIOR_LAST
    3     MODE_INTER_NO_MV
    4     MODE_INTRA
    5     MODE_USING_GOLDEN
    6     MODE_GOLDEN_MV
    7     MODE_INTER_FOURMV
    
Alphabet 4:
  index   coding mode
  -----   -----------
    0     MODE_INTER_LAST_MV
    1     MODE_INTER_PLUS_MV
    2     MODE_INTER_NO_MV
    3     MODE_INTER_PRIOR_LAST
    4     MODE_INTRA
    5     MODE_USING_GOLDEN
    6     MODE_GOLDEN_MV
    7     MODE_INTER_FOURMV

Alphabet 5:
  index   coding mode
  -----   -----------
    0     MODE_INTER_NO_MV
    1     MODE_INTER_LAST_MV
    2     MODE_INTER_PRIOR_LAST
    3     MODE_INTER_PLUS_MV
    4     MODE_INTRA
    5     MODE_USING_GOLDEN
    6     MODE_GOLDEN_MV
    7     MODE_INTER_FOURMV
    
Alphabet 6:
  index   coding mode
  -----   -----------
    0     MODE_INTER_NO_MV
    1     MODE_USING_GOLDEN
    2     MODE_INTER_LAST_MV
    3     MODE_INTER_PRIOR_LAST
    4     MODE_INTER_PLUS_MV
    5     MODE_INTRA
    6     MODE_GOLDEN_MV
    7     MODE_INTER_FOURMV


Appendix C: DCT Coefficient VLC Tables
--------------------------------------
- VP31 tables are hardcoded
- Theora tables are transported with video stream

[not finished]


Appendix D: The VP3 IDCT
------------------------

[not finished]


Acknowledgements
----------------
Thanks to Michael Niedermayer (michaelni at gmx dot at) for peer review,
corrections, and recommendations for improvement.

Dan Miller (dan at on2 dot com) for clarifications on pieces of the
format.

Timothy B. Terriberry (tterribe at vt dot edu) for clarification about the
differences between VP3 and Theora, detailed explanation of motion 
vector mechanics.


References
----------
Tables necessary for decoding VP3:
http://mplayerhq.hu/cgi-bin/cvsweb.cgi/~checkout~/ffmpeg/libavcodec/vp3data.h?content-type=text/x-cvsweb-markup&cvsroot=FFMpeg

Official VP3 site:
http://www.vp3.com/

Theora, based on VP3:
http://www.theora.org/

On2, creators of the VP3 format:
http://www.on2.com/


ChangeLog
---------
v0.5: December 8, 2004
- reworked section "Reversing The DC Prediction" to include a tabular 
representation of all 16 prediction modes

v0.4: March 2, 2004
- renamed and expanded section "Initializing The Quantization Matrices"
- outlined section "Reconstructing The Frame"
- moved Theora Differences Appendix to its own section entitled "Theora
Specification"
- added Appendix: Quantization Matrices And Scale Factors
- added Appendix: DCT Coefficient VLC Tables

v0.3: February 29, 2004
- expanded section "Unpacking The Macroblock Coding Mode Information"
- expanded section "Unpacking The Macroblock Motion Vectors"
- added Appendix: Macroblock Coding Mode Alphabets

v0.2: October 9, 2003
- expanded section "Reversing the DC Prediction"
- added Appendix: Theora Differences

v0.1: June 17, 2003
- initial release, nowhere near complete