H.264 Prediction: Difference between revisions

From MultimediaWiki
Jump to navigation Jump to search
m (→‎Plane: ohh, Mike, Mike...)
 
(36 intermediate revisions by 4 users not shown)
Line 1: Line 1:
This page documents the various prediction methods used in [[H.264]] and related formats such as [[Sorenson Video 3]] and [[RealVideo 4]].
This page documents the various prediction methods used in [[H.264]] and related formats such as [[Sorenson Video 3]], [[RealVideo 4]], and [[On2 VP8]].


== 4x4 Prediction Modes ==
== 4x4 Prediction Modes ==
4x4 prediction modes vary between different codecs. While they are almost the same for [[H.264]] and [[Sorenson Video 3]], [[RealVideo 4]] has a different order for these modes and some of them significantly differ from H.264 counterparts (by using left predictors where H.264 does not and down left predictors which are not used elsewhere).


=== Vertical ===
=== Vertical ===
Line 7: Line 9:
* H.264: mode 0
* H.264: mode 0
* SVQ3: mode 0
* SVQ3: mode 0
* RV40: mode 0
* RV40: mode 1
* VP8: not used


  LT | T0  T1  T2  T3
    | T0  T1  T2  T3
  ---------------------
  ---------------------
  L0 | T0  T1  T2  T3
    | T0  T1  T2  T3
  L1 | T0  T1  T2  T3
    | T0  T1  T2  T3
  L2 | T0  T1  T2  T3
    | T0  T1  T2  T3
   L3 | T0  T1  T2  T3
    | T0  T1  T2  T3
 
=== Vertical (VP8) ===
 
* H.264: not used
* SVQ3: not used
* RV40: not used
* VP8: mode 2
 
   LT | T0  T1  T2  T3 T4
------------------------
    |  a  b  c  d
    |  a  b  c  d
    |  a  b  c  d
    |  a  b  c  d
 
where:
 
  a = (LT + 2*T0 + T1 + 2) >> 2
  b = (T0 + 2*T1 + T2 + 2) >> 2
  c = (T1 + 2*T2 + T3 + 2) >> 2
  d = (T2 + 2*T3 + T4 + 2) >> 2


=== Horizontal ===
=== Horizontal ===
Line 20: Line 44:
* H.264: mode 1
* H.264: mode 1
* SVQ3: mode 1
* SVQ3: mode 1
* RV40: mode 1
* RV40: mode 2
* VP8: not used


  LT | T0  T1  T2  T3
    |  
  ---------------------
  ---------------------
   L0 | L0  L0  L0  L0
   L0 | L0  L0  L0  L0
Line 28: Line 53:
   L2 | L2  L2  L2  L2
   L2 | L2  L2  L2  L2
   L3 | L3  L3  L3  L3
   L3 | L3  L3  L3  L3
=== Horizontal (VP8) ===
* H.264: not used
* SVQ3: not used
* RV40: not used
* VP8: mode 3
  LT |
--------------------
  L0 |  a  a  a  a
  L1 |  b  b  b  b
  L2 |  c  c  c  c
  L3 |  d  d  d  d
where:
  a = (LT + 2*L0 + L1 + 2) >> 2
  b = (L0 + 2*L1 + L2 + 2) >> 2
  c = (L1 + 2*L2 + L3 + 2) >> 2
  d = (L2 + 2*L3 + L3 + 2) >> 2


=== DC ===
=== DC ===
Line 33: Line 79:
* H.264: mode 2
* H.264: mode 2
* SVQ3: mode 2
* SVQ3: mode 2
* RV40: mode 2
* RV40: mode 0
* VP8: mode 0


  LT | T0  T1  T2  T3
    | T0  T1  T2  T3
  ---------------------
  ---------------------
   L0 |  a  a  a  a
   L0 |  a  a  a  a
Line 44: Line 91:
where:
where:


  a = (T0 + T1 + T2 + T3 + L0 + L1 + L2 + L3 + 4) / 8
  if top and left predictors are available
  a = (T0 + T1 + T2 + T3 + L0 + L1 + L2 + L3 + 4) / 8
else if top predictors are available
  a = (T0 + T1 + T2 + T3 + 2) / 4
else if left predictors are available
  a = (L0 + L1 + L2 + L3 + 2) / 4
else
  a = 128
 
Note that the VP8 reference code does not make any provisions for either or both sets of predictors to be missing.


=== Diagonal Down/Left ===
=== Diagonal Down/Left ===
Line 50: Line 106:
* H.264: mode 3
* H.264: mode 3
* SVQ3: not used
* SVQ3: not used
* RV40: mode 3
* RV40: not used
* VP8: mode 4


  LT | T0  T1  T2  T3  T4  T5  T6  T7
    | T0  T1  T2  T3  T4  T5  T6  T7
  -------------------------------------
  -------------------------------------
  L0 |  a  b  c  d
    |  a  b  c  d
  L1 |  b  c  d  e
    |  b  c  d  e
  L2 |  c  d  e  f
    |  c  d  e  f
  L3 |  d  e  f  g
    |  d  e  f  g


where:
where:
Line 67: Line 124:
   e = (T4 + 2*T5 + T6 + 2) / 4
   e = (T4 + 2*T5 + T6 + 2) / 4
   f = (T5 + 2*T6 + T7 + 2) / 4
   f = (T5 + 2*T6 + T7 + 2) / 4
   g = (T6 * 3*T7      + 2) / 4
   g = (T6 + 3*T7      + 2) / 4


=== Diagonal Down/Left (SVQ3) ===
=== Diagonal Down/Left (SVQ3) ===
Line 74: Line 131:
* SVQ3: mode 3
* SVQ3: mode 3
* RV40: not used
* RV40: not used
* VP8: not used


  LT | T0  T1  T2  T3
    |     T1  T2  T3
  ---------------------
  ---------------------
  L0 |  a  b  c  c
    |  a  b  c  c
   L1 |  b  c  c  c
   L1 |  b  c  c  c
   L2 |  c  c  c  c
   L2 |  c  c  c  c
Line 87: Line 145:
   b = (L2 + T2) / 2
   b = (L2 + T2) / 2
   c = (L3 + T3) / 2
   c = (L3 + T3) / 2
=== Diagonal Down/Left (RV40) ===
* H.264: not used
* SVQ3: not used
* RV40: mode 4
* VP8: not used
    | T0  T1  T2  T3  T4  T5  T6  T7
-------------------------------------
  L0 |  a  b  c  d
  L1 |  b  c  d  e
  L2 |  c  d  e  f
  L3 |  d  e  f  g
  L4 |
  L5 |
  L6 |
  L7 |
where:
  a = (T0 + 2*T1 + T2 + L0 + 2*L1 + L2 + 4) / 8
  b = (T1 + 2*T2 + T3 + L1 + 2*L2 + L3 + 4) / 8
  c = (T2 + 2*T3 + T4 + L2 + 2*L3 + L4 + 4) / 8
  d = (T3 + 2*T4 + T5 + L3 + 2*L4 + L5 + 4) / 8
  e = (T4 + 2*T5 + T6 + L4 + 2*L5 + L6 + 4) / 8
  f = (T5 + 2*T6 + T7 + L5 + 2*L6 + L7 + 4) / 8
  g = (T6 +  T7      + L6 +  L7      + 2) / 4


=== Diagonal Down/Right ===
=== Diagonal Down/Right ===
Line 92: Line 178:
* H.264: mode 4
* H.264: mode 4
* SVQ3: mode 4
* SVQ3: mode 4
* RV40: mode 4
* RV40: mode 3
* VP8: mode 5


   LT | T0  T1  T2  T3
   LT | T0  T1  T2  T3
Line 116: Line 203:
* SVQ3: mode 5
* SVQ3: mode 5
* RV40: mode 5
* RV40: mode 5
* VP8: mode 6


   LT | T0  T1  T2  T3
   LT | T0  T1  T2  T3
Line 122: Line 210:
   L1 |  e  f  g  h
   L1 |  e  f  g  h
   L2 |  i  a  b  c
   L2 |  i  a  b  c
  L3 |  j  e  f  g
    |  j  e  f  g


where:
where:
Line 141: Line 229:
* H.264: mode 6
* H.264: mode 6
* SVQ3: mode 6
* SVQ3: mode 6
* RV40: mode 6
* RV40: mode 8
* VP8: mode 8


   LT | T0  T1  T2  T3
   LT | T0  T1  T2   
  ---------------------
  ---------------------
   L0 |  a  b  c  d
   L0 |  a  b  c  d
Line 167: Line 256:
* H.264: mode 7
* H.264: mode 7
* SVQ3: mode 7
* SVQ3: mode 7
* RV40: mode 7
* RV40: mode 6
* VP8: mode 7


  LT | T0  T1  T2  T3  T4  T5  T6 T7
    | T0  T1  T2  T3  T4  T5  T6  
  -------------------------------------
  ---------------------------------
  L0 |  a  b  c  d
    |  a  b  c  d
   L1 |  f  g  h  i
   L1 |  f  g  h  i
   L2 |  b  c  d  e
   L2 |  b  c  d  e
   L3 |  g  h  i  j
   L3 |  g  h  i  j
  L4 |


where:
where:
Line 188: Line 279:
   i = (T3 + 2*T4 + T5 + 2) / 4
   i = (T3 + 2*T4 + T5 + 2) / 4
   j = (T4 + 2*T5 + T6 + 2) / 4
   j = (T4 + 2*T5 + T6 + 2) / 4
For RV40 two coefficients differ:
  a = (2*T0 + 2*T1 + L1 + 2*L2 +  L3 +      4) / 8
  f = (  T0 + 2*T1 + T2 +  L2 + 2*L3 + L4 + 4) / 8
For VP8, 3 coefficients differ:
  c = (T2 + T3 + T4 + 2) / 4
  e = (T4 + 2*T5 + T6 + 2) / 4
  j = (T5 + 2*T6 + T7 + 2) / 4


=== Horizontal/Up ===
=== Horizontal/Up ===
Line 193: Line 295:
* H.264: mode 8
* H.264: mode 8
* SVQ3: mode 8
* SVQ3: mode 8
* RV40: mode 8
* RV40: not used
* VP8: mode 9


  LT | T0  T1  T2  T3
    |  
  ---------------------
  ---------------------
   L0 |  a  b  c  d
   L0 |  a  b  c  d
Line 209: Line 312:
   d = (L1 + 2*L2 + L3 + 2) / 4
   d = (L1 + 2*L2 + L3 + 2) / 4
   e = (L2 + L3 + 1) / 2
   e = (L2 + L3 + 1) / 2
   f = (L2 + 2*L3 + L3 + 2) / 4
   f = (L2 + 3*L3     + 2) / 4
   g = L3
   g = L3


=== Left/DC ===
=== Horizontal/Up (RV40)===


* H.264: mode 9
* H.264: not used
* SVQ3: mode 9
* SVQ3: not used
* RV40: mode 9
* RV40: mode 7
* VP8: not used


  LT | T0  T1  T2  T3
    | T0  T1  T2  T3 T4  T5  T6  T7
  ---------------------
  -------------------------------------
   L0 |  a  a   a   a
   L0 |  a  b   c   d
   L1 |  a   a   a   a
   L1 |  c   d   e   f
   L2 |  a   a   a   a
   L2 |  e   f   g   h
   L3 |  a   a   a   a
   L3 |  g  h  i  j
   L4 |
   L5 |
   L6 |


where:
where:


a = (L0 + L1 + L2 + L3 + 2) / 4
  a = (T1 + 2*T2 + T3 + 2*L0 + 2*L1 +      4) / 8
  b = (T2 + 2*T3 + T4 +  L0 + 2*L1 + L2 + 4) / 8
  c = (T3 + 2*T4 + T5 + 2*L1 + 2*L2 +      4) / 8
  d = (T4 + 2*T5 + T6 +  L1 + 2*L2 + L3 + 4) / 8
  e = (T5 + 2*T6 + T7 + 2*L2 + 2*L3 +      4) / 8
  f = (T6 + 3*T7 +        L2 + 3*L3 +      4) / 8
  g = (T6 +  T7 +        L3 +  L4      + 2) / 4
  h = (                  L3 + 2*L4 + L5 + 2) / 4
  i = (                  L4 +  L5      + 1) / 2
  j = (                  L4 + 2*L5 + L6 + 2) / 4


=== Top/DC ===
=== TrueMotion (VP8) ===


* H.264: mode 10
* H.264: not used
* SVQ3: mode 10
* SVQ3: not used
* RV40: mode 10
* RV40: not used
* VP8: mode 1


   LT | T0  T1  T2  T3
   LT | T0  T1  T2  T3
  ---------------------
  ---------------------
   L0 |  a  a   a   a
   L0 |  a  .   .   .
   L1 |  a   a   a   a
   L1 |  .   b   .   .
   L2 |  a   a   a   a
   L2 |  .   .   c   .
   L3 |  a   a   a   a
   L3 |  .   .   .   d


where:
where this pattern is satisfied:


a = (T0 + T1 + T2 + T3 + 2) / 4
  a = SATURATE_U8(T0 - LT + L0)
  b = SATURATE_U8(T1 - LT + L1)
  c = SATURATE_U8(T2 - LT + L2)
  d = SATURATE_U8(T3 - LT + L3)


=== DC-128 ===
I.e., for each of the 16 samples: (top predictor for column) - (left/top predictor) + (left predictor for row), then saturate in an unsigned byte range 0..255.


* H.264: mode 11
* SVQ3: mode 11
* RV40: mode 11
  LT |  T0  T1  T2  T3
------------------------
  L0 | 128  128  128  128
  L1 | 128  128  128  128
  L2 | 128  128  128  128
  L3 | 128  128  128  128


== 16x16 Prediction Modes ==
== 16x16 Prediction Modes ==
Line 266: Line 376:
* SVQ3: mode 0
* SVQ3: mode 0
* RV40: mode 0
* RV40: mode 0
* VP8: mode 0


Using the 16 top predictors (T0..T15) and the 16 left predictors (L0..L15), set all 256 elements to the mean, computed as:
Using the 16 top predictors (T0..T15) and the 16 left predictors (L0..L15), set all 256 elements to the mean, computed as:


  mean = (sum(T0..T15) + sum(L0..L15) + 16) / 32
if top and left predictors are available
  mean = (sum(T0..T15) + sum(L0..L15) + 16) / 32
else if top predictors are available
  mean = (sum(T0..T15) + 8) / 16
else if left predictors are available
  mean = (sum(L0..L15) + 8) / 16
else
  mean = 128


=== Vertical ===
=== Vertical ===
Line 276: Line 394:
* SVQ3: mode 1
* SVQ3: mode 1
* RV40: mode 1
* RV40: mode 1
* VP8: mode 1


  LT | T0  T1  T2  T3  T4  ..  T15
      | T0  T1  T2  T3  T4  ..  T15
  ------------------------- .. -----
  -------------------------- .. -----
  L0 | T0  T1  T2  T3  T4  ..  T15
      | T0  T1  T2  T3  T4  ..  T15
  L1 | T0  T1  T2  T3  T4  ..  T15
      | T0  T1  T2  T3  T4  ..  T15
  L2 | T0  T1  T2  T3  T4  ..  T15
      | T0  T1  T2  T3  T4  ..  T15
   ......
   ......
  L15 | T0  T1  T2  T3  T4  ..  T15
      | T0  T1  T2  T3  T4  ..  T15


=== Horizontal ===
=== Horizontal ===
Line 290: Line 409:
* SVQ3: mode 2
* SVQ3: mode 2
* RV40: mode 2
* RV40: mode 2
* VP8: mode 2


  LT | T0  T1  T2  T3  T4  ..  T15
      |
  --------------------------- .. -----
  --------------------------- .. -----
   L0 |  L0  L0  L0  L0  L0  ..  L0
   L0 |  L0  L0  L0  L0  L0  ..  L0
Line 304: Line 424:
* SVQ3: mode 3
* SVQ3: mode 3
* RV40: mode 3
* RV40: mode 3
* VP8: not used
Notice that SVQ3 and RV40 follow a slightly different method here.
Given the top predictors (T0..T15), left predictors (L0..L15) and the left-top corner predictor (LT) arranged as follows:
  LT |    T0        T1        T2    ..    T15
-----------------------------------  ..  --------
  L0 | c[ 0, 0]  c[ 1, 0]  c[ 2, 0]  ..  c[15, 0]
  L1 | c[ 0, 1]  c[ 1, 1]  c[ 2, 1]  ..  c[15, 1]
  ......
  L15 | c[ 0,15]  c[ 1,15]  c[ 2,15]  ..  c[15,15]
Compute H' and V':
H' = 1* (T8 - T6) +
      2* (T9 - T5) +
      3*(T10 - T4) +
      4*(T11 - T3) +
      5*(T12 - T2) +
      6*(T13 - T1) +
      7*(T14 - T0) +
      8*(T15 - LT)
V' = 1* (L8 - L6) +
      2* (L9 - L5) +
      3*(L10 - L4) +
      4*(L11 - L3) +
      5*(L12 - L2) +
      6*(L13 - L1) +
      7*(L14 - L0) +
      8*(L15 - LT)
For H.264, compute H and V as:
  H = (5*H' + 32) / 64
  V = (5*V' + 32) / 64
For SVQ3, compute H and V as:
  V = (5*(H'/4)) / 16
  H = (5*(V'/4)) / 16
  (notice that V and H are computed from H' and V', respectively)
For RV40, compute H and V as:
  H = (5*(H' >> 2)) >> 4
  V = (5*(V' >> 2)) >> 4
  (like SVQ3 but without swapping and it's important to use shifts here instead of divisions)
The final process for filling in the 16x16 block is:
  a = 16 * (L15 + T15 + 1) - 7*(V+H)
  for (j = 0..15)
    for (i = 0..15)
      b = a + V * j + H * i
      c[i,j] = SATURATE_U8(b / 32)
The SATURATE_U8() function indicates that the result of the operation should be bounded to an unsigned 8-bit range (0..255).


''SVQ3 follows a slightly different method here''
=== TrueMotion (VP8) ===


'''(not done yet)'''
* H.264: not used
* SVQ3: not used
* RV40: not used
* VP8: mode 3
 
Given the top predictors (T0..T15), left predictors (L0..L15) and the left-top corner predictor (LT) arranged as follows:
 
  LT |    T0        T1        T2    ..    T15
-----------------------------------  ..  --------
  L0 | c[ 0, 0]  c[ 1, 0]  c[ 2, 0]  ..  c[15, 0]
  L1 | c[ 0, 1]  c[ 1, 1]  c[ 2, 1]  ..  c[15, 1]
  ......
  L15 | c[ 0,15]  c[ 1,15]  c[ 2,15]  ..  c[15,15]


=== Left/DC ===
Compute c[t,l] as:


=== Top/DC ===
  c[t,l] = SATURATE_U8(L[l] + T[t] - LT)


=== DC-128 ===
I.e., for each element, sum the left and top predictors for the row and column, respectively, and subtract the left-top predictor. Then, saturate the result between 0 and 255.


[[Category:Compression Theory]]
[[Category:Compression Theory]]

Latest revision as of 23:50, 22 April 2011

This page documents the various prediction methods used in H.264 and related formats such as Sorenson Video 3, RealVideo 4, and On2 VP8.

4x4 Prediction Modes

4x4 prediction modes vary between different codecs. While they are almost the same for H.264 and Sorenson Video 3, RealVideo 4 has a different order for these modes and some of them significantly differ from H.264 counterparts (by using left predictors where H.264 does not and down left predictors which are not used elsewhere).

Vertical

  • H.264: mode 0
  • SVQ3: mode 0
  • RV40: mode 1
  • VP8: not used
    | T0  T1  T2  T3
---------------------
    | T0  T1  T2  T3
    | T0  T1  T2  T3
    | T0  T1  T2  T3
    | T0  T1  T2  T3

Vertical (VP8)

  • H.264: not used
  • SVQ3: not used
  • RV40: not used
  • VP8: mode 2
 LT | T0  T1  T2  T3  T4
------------------------
    |  a   b   c   d
    |  a   b   c   d
    |  a   b   c   d
    |  a   b   c   d

where:

 a = (LT + 2*T0 + T1 + 2) >> 2
 b = (T0 + 2*T1 + T2 + 2) >> 2
 c = (T1 + 2*T2 + T3 + 2) >> 2
 d = (T2 + 2*T3 + T4 + 2) >> 2

Horizontal

  • H.264: mode 1
  • SVQ3: mode 1
  • RV40: mode 2
  • VP8: not used
    | 
---------------------
 L0 | L0  L0  L0  L0
 L1 | L1  L1  L1  L1
 L2 | L2  L2  L2  L2
 L3 | L3  L3  L3  L3

Horizontal (VP8)

  • H.264: not used
  • SVQ3: not used
  • RV40: not used
  • VP8: mode 3
 LT | 
--------------------
 L0 |  a   a   a   a
 L1 |  b   b   b   b
 L2 |  c   c   c   c
 L3 |  d   d   d   d

where:

 a = (LT + 2*L0 + L1 + 2) >> 2
 b = (L0 + 2*L1 + L2 + 2) >> 2
 c = (L1 + 2*L2 + L3 + 2) >> 2
 d = (L2 + 2*L3 + L3 + 2) >> 2

DC

  • H.264: mode 2
  • SVQ3: mode 2
  • RV40: mode 0
  • VP8: mode 0
    | T0  T1  T2  T3
---------------------
 L0 |  a   a   a   a
 L1 |  a   a   a   a
 L2 |  a   a   a   a
 L3 |  a   a   a   a

where:

if top and left predictors are available
  a = (T0 + T1 + T2 + T3 + L0 + L1 + L2 + L3 + 4) / 8
else if top predictors are available
  a = (T0 + T1 + T2 + T3 + 2) / 4
else if left predictors are available
  a = (L0 + L1 + L2 + L3 + 2) / 4
else
  a = 128

Note that the VP8 reference code does not make any provisions for either or both sets of predictors to be missing.

Diagonal Down/Left

  • H.264: mode 3
  • SVQ3: not used
  • RV40: not used
  • VP8: mode 4
    | T0  T1  T2  T3  T4  T5  T6  T7
-------------------------------------
    |  a   b   c   d
    |  b   c   d   e
    |  c   d   e   f
    |  d   e   f   g

where:

 a = (T0 + 2*T1 + T2 + 2) / 4
 b = (T1 + 2*T2 + T3 + 2) / 4
 c = (T2 + 2*T3 + T4 + 2) / 4
 d = (T3 + 2*T4 + T5 + 2) / 4
 e = (T4 + 2*T5 + T6 + 2) / 4
 f = (T5 + 2*T6 + T7 + 2) / 4
 g = (T6 + 3*T7      + 2) / 4

Diagonal Down/Left (SVQ3)

  • H.264: not used
  • SVQ3: mode 3
  • RV40: not used
  • VP8: not used
    |     T1  T2  T3
---------------------
    |  a   b   c   c
 L1 |  b   c   c   c
 L2 |  c   c   c   c
 L3 |  c   c   c   c

where:

 a = (L1 + T1) / 2
 b = (L2 + T2) / 2
 c = (L3 + T3) / 2

Diagonal Down/Left (RV40)

  • H.264: not used
  • SVQ3: not used
  • RV40: mode 4
  • VP8: not used
    | T0  T1  T2  T3  T4  T5  T6  T7
-------------------------------------
 L0 |  a   b   c   d
 L1 |  b   c   d   e
 L2 |  c   d   e   f
 L3 |  d   e   f   g
 L4 |
 L5 |
 L6 |
 L7 |

where:

 a = (T0 + 2*T1 + T2 + L0 + 2*L1 + L2 + 4) / 8
 b = (T1 + 2*T2 + T3 + L1 + 2*L2 + L3 + 4) / 8
 c = (T2 + 2*T3 + T4 + L2 + 2*L3 + L4 + 4) / 8
 d = (T3 + 2*T4 + T5 + L3 + 2*L4 + L5 + 4) / 8
 e = (T4 + 2*T5 + T6 + L4 + 2*L5 + L6 + 4) / 8
 f = (T5 + 2*T6 + T7 + L5 + 2*L6 + L7 + 4) / 8
 g = (T6 +   T7      + L6 +   L7      + 2) / 4

Diagonal Down/Right

  • H.264: mode 4
  • SVQ3: mode 4
  • RV40: mode 3
  • VP8: mode 5
 LT | T0  T1  T2  T3
---------------------
 L0 |  d   e   f   g
 L1 |  c   d   e   f
 L2 |  b   c   d   e
 L3 |  a   b   c   d

where:

 a = (L3 + 2*L2 + L1 + 2) / 4
 b = (L2 + 2*L1 + L0 + 2) / 4
 c = (L1 + 2*L0 + LT + 2) / 4
 d = (L0 + 2*LT + T0 + 2) / 4
 e = (LT + 2*T0 + T1 + 2) / 4
 f = (T0 + 2*T1 + T2 + 2) / 4
 g = (T1 + 2*T2 + T3 + 2) / 4

Vertical/Right

  • H.264: mode 5
  • SVQ3: mode 5
  • RV40: mode 5
  • VP8: mode 6
 LT | T0  T1  T2  T3
---------------------
 L0 |  a   b   c   d
 L1 |  e   f   g   h
 L2 |  i   a   b   c
    |  j   e   f   g

where:

 a = (LT + T0 + 1) / 2
 b = (T0 + T1 + 1) / 2
 c = (T1 + T2 + 1) / 2
 d = (T2 + T3 + 1) / 2
 e = (L0 + 2*LT + T0 + 2) / 4
 f = (LT + 2*T0 + T1 + 2) / 4
 g = (T0 + 2*T1 + T2 + 2) / 4
 h = (T1 + 2*T2 + T3 + 2) / 4
 i = (LT + 2*L0 + L1 + 2) / 4
 j = (L0 + 2*L1 + L2 + 2) / 4

Horizontal/Down

  • H.264: mode 6
  • SVQ3: mode 6
  • RV40: mode 8
  • VP8: mode 8
 LT | T0  T1  T2  
---------------------
 L0 |  a   b   c   d
 L1 |  e   f   a   b
 L2 |  g   h   e   f
 L3 |  i   j   g   h

where:

 a = (LT + L0 + 1) / 2
 b = (L0 + 2*LT + T0 + 2) / 4
 c = (LT + 2*T0 + T1 + 2) / 4
 d = (T0 + 2*T1 + T2 + 2) / 4
 e = (L0 + L1 + 1) / 2
 f = (LT + 2*L0 + L1 + 2) / 4
 g = (L1 + L2 + 1) / 2
 h = (L0 + 2*L1 + L2 + 2) / 4
 g = (L2 + L3 + 1) / 2
 j = (L1 + 2*L2 + L3 + 2) / 4

Vertical/Left

  • H.264: mode 7
  • SVQ3: mode 7
  • RV40: mode 6
  • VP8: mode 7
    | T0  T1  T2  T3  T4  T5  T6 
---------------------------------
    |  a   b   c   d
 L1 |  f   g   h   i
 L2 |  b   c   d   e
 L3 |  g   h   i   j
 L4 |

where:

 a = (T0 + T1 + 1) / 2
 b = (T1 + T2 + 1) / 2
 c = (T2 + T3 + 1) / 2
 d = (T3 + T4 + 1) / 2
 e = (T4 + T5 + 1) / 2
 f = (T0 + 2*T1 + T2 + 2) / 4
 g = (T1 + 2*T2 + T3 + 2) / 4
 h = (T2 + 2*T3 + T4 + 2) / 4
 i = (T3 + 2*T4 + T5 + 2) / 4
 j = (T4 + 2*T5 + T6 + 2) / 4

For RV40 two coefficients differ:

 a = (2*T0 + 2*T1 + L1 + 2*L2 +   L3 +      4) / 8
 f = (  T0 + 2*T1 + T2 +   L2 + 2*L3 + L4 + 4) / 8

For VP8, 3 coefficients differ:

 c = (T2 + T3 + T4 + 2) / 4
 e = (T4 + 2*T5 + T6 + 2) / 4
 j = (T5 + 2*T6 + T7 + 2) / 4

Horizontal/Up

  • H.264: mode 8
  • SVQ3: mode 8
  • RV40: not used
  • VP8: mode 9
    | 
---------------------
 L0 |  a   b   c   d
 L1 |  c   d   e   f
 L2 |  e   f   g   g
 L3 |  g   g   g   g

where:

 a = (L0 + L1 + 1) / 2
 b = (L0 + 2*L1 + L2 + 2) / 4
 c = (L1 + L2 + 1) / 2
 d = (L1 + 2*L2 + L3 + 2) / 4
 e = (L2 + L3 + 1) / 2
 f = (L2 + 3*L3      + 2) / 4
 g = L3

Horizontal/Up (RV40)

  • H.264: not used
  • SVQ3: not used
  • RV40: mode 7
  • VP8: not used
    | T0  T1  T2  T3  T4  T5  T6  T7
-------------------------------------
 L0 |  a   b   c   d
 L1 |  c   d   e   f
 L2 |  e   f   g   h
 L3 |  g   h   i   j
 L4 |
 L5 |
 L6 |

where:

 a = (T1 + 2*T2 + T3 + 2*L0 + 2*L1 +      4) / 8
 b = (T2 + 2*T3 + T4 +   L0 + 2*L1 + L2 + 4) / 8
 c = (T3 + 2*T4 + T5 + 2*L1 + 2*L2 +      4) / 8
 d = (T4 + 2*T5 + T6 +   L1 + 2*L2 + L3 + 4) / 8
 e = (T5 + 2*T6 + T7 + 2*L2 + 2*L3 +      4) / 8
 f = (T6 + 3*T7 +        L2 + 3*L3 +      4) / 8
 g = (T6 +   T7 +        L3 +   L4      + 2) / 4
 h = (                   L3 + 2*L4 + L5 + 2) / 4
 i = (                   L4 +   L5      + 1) / 2
 j = (                   L4 + 2*L5 + L6 + 2) / 4

TrueMotion (VP8)

  • H.264: not used
  • SVQ3: not used
  • RV40: not used
  • VP8: mode 1
 LT | T0  T1  T2  T3
---------------------
 L0 |  a   .   .   .
 L1 |  .   b   .   .
 L2 |  .   .   c   .
 L3 |  .   .   .   d

where this pattern is satisfied:

 a = SATURATE_U8(T0 - LT + L0)
 b = SATURATE_U8(T1 - LT + L1)
 c = SATURATE_U8(T2 - LT + L2)
 d = SATURATE_U8(T3 - LT + L3)

I.e., for each of the 16 samples: (top predictor for column) - (left/top predictor) + (left predictor for row), then saturate in an unsigned byte range 0..255.


16x16 Prediction Modes

DC

  • H.264: mode 0
  • SVQ3: mode 0
  • RV40: mode 0
  • VP8: mode 0

Using the 16 top predictors (T0..T15) and the 16 left predictors (L0..L15), set all 256 elements to the mean, computed as:

if top and left predictors are available
  mean = (sum(T0..T15) + sum(L0..L15) + 16) / 32
else if top predictors are available
  mean = (sum(T0..T15) + 8) / 16
else if left predictors are available
  mean = (sum(L0..L15) + 8) / 16
else
  mean = 128

Vertical

  • H.264: mode 1
  • SVQ3: mode 1
  • RV40: mode 1
  • VP8: mode 1
     | T0  T1  T2  T3  T4  ..  T15
-------------------------- .. -----
     | T0  T1  T2  T3  T4  ..  T15
     | T0  T1  T2  T3  T4  ..  T15
     | T0  T1  T2  T3  T4  ..  T15
 ......
     | T0  T1  T2  T3  T4  ..  T15

Horizontal

  • H.264: mode 2
  • SVQ3: mode 2
  • RV40: mode 2
  • VP8: mode 2
     |
--------------------------- .. -----
  L0 |  L0  L0  L0  L0  L0  ..   L0
  L1 |  L1  L1  L1  L1  L1  ..   L1
  L2 |  L2  L2  L2  L2  L2  ..   L2
 ......
 L15 | L15 L15 L15 L15 L15  ..  L15

Plane

  • H.264: mode 3
  • SVQ3: mode 3
  • RV40: mode 3
  • VP8: not used

Notice that SVQ3 and RV40 follow a slightly different method here.

Given the top predictors (T0..T15), left predictors (L0..L15) and the left-top corner predictor (LT) arranged as follows:

  LT |    T0        T1        T2     ..    T15
-----------------------------------  ..  --------
  L0 | c[ 0, 0]  c[ 1, 0]  c[ 2, 0]  ..  c[15, 0]
  L1 | c[ 0, 1]  c[ 1, 1]  c[ 2, 1]  ..  c[15, 1]
 ......
 L15 | c[ 0,15]  c[ 1,15]  c[ 2,15]  ..  c[15,15]

Compute H' and V':

H' = 1* (T8 - T6) +
     2* (T9 - T5) +
     3*(T10 - T4) +
     4*(T11 - T3) +
     5*(T12 - T2) +
     6*(T13 - T1) +
     7*(T14 - T0) +
     8*(T15 - LT)
V' = 1* (L8 - L6) +
     2* (L9 - L5) +
     3*(L10 - L4) +
     4*(L11 - L3) +
     5*(L12 - L2) +
     6*(L13 - L1) +
     7*(L14 - L0) +
     8*(L15 - LT)

For H.264, compute H and V as:

 H = (5*H' + 32) / 64
 V = (5*V' + 32) / 64

For SVQ3, compute H and V as:

 V = (5*(H'/4)) / 16
 H = (5*(V'/4)) / 16 
 (notice that V and H are computed from H' and V', respectively)

For RV40, compute H and V as:

 H = (5*(H' >> 2)) >> 4
 V = (5*(V' >> 2)) >> 4 
 (like SVQ3 but without swapping and it's important to use shifts here instead of divisions)

The final process for filling in the 16x16 block is:

 a = 16 * (L15 + T15 + 1) - 7*(V+H)
 for (j = 0..15)
   for (i = 0..15)
     b = a + V * j + H * i
     c[i,j] = SATURATE_U8(b / 32)

The SATURATE_U8() function indicates that the result of the operation should be bounded to an unsigned 8-bit range (0..255).

TrueMotion (VP8)

  • H.264: not used
  • SVQ3: not used
  • RV40: not used
  • VP8: mode 3

Given the top predictors (T0..T15), left predictors (L0..L15) and the left-top corner predictor (LT) arranged as follows:

  LT |    T0        T1        T2     ..    T15
-----------------------------------  ..  --------
  L0 | c[ 0, 0]  c[ 1, 0]  c[ 2, 0]  ..  c[15, 0]
  L1 | c[ 0, 1]  c[ 1, 1]  c[ 2, 1]  ..  c[15, 1]
 ......
 L15 | c[ 0,15]  c[ 1,15]  c[ 2,15]  ..  c[15,15]

Compute c[t,l] as:

 c[t,l] = SATURATE_U8(L[l] + T[t] - LT)

I.e., for each element, sum the left and top predictors for the row and column, respectively, and subtract the left-top predictor. Then, saturate the result between 0 and 255.