blob: d7549dcc658d9b5f74148f27f9e9c567bdafff72 [file] [log] [blame]
#### 8.2 Tree Coding Example {#h-08-02}
As a multi-part example, without getting too far into the semantics of macroblock decoding (which is of course taken up below), we look at the "mode" coding for intra-predicted macroblocks.
It so happens that, because of a difference in statistics, the Y (or luma) mode encoding uses two different trees: one for key frames and another for interframes. This is the only instance in VP8 of the same dataset being coded by different trees under different circumstances. The UV (or chroma) modes are a proper subset of the Y modes and, as such, have their own decoding tree.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
typedef enum
{
DC_PRED, /* predict DC using row above and column to the left */
V_PRED, /* predict rows using row above */
H_PRED, /* predict columns using column to the left */
TM_PRED, /* propagate second differences a la "True Motion" */
B_PRED, /* each Y subblock is independently predicted */
num_uv_modes = B_PRED, /* first four modes apply to chroma */
num_ymodes /* all modes apply to luma */
}
intra_mbmode;
/* The aforementioned trees together with the implied codings as
comments.
Actual (i.e., positive) indices are always even.
Value (i.e., nonpositive) indices are arbitrary. */
const tree_index ymode_tree [2 * (num_ymodes - 1)] =
{
-DC_PRED, 2, /* root: DC_PRED = "0", "1" subtree */
4, 6, /* "1" subtree has 2 descendant subtrees */
-V_PRED, -H_PRED, /* "10" subtree: V_PRED = "100",
H_PRED = "101" */
-TM_PRED, -B_PRED /* "11" subtree: TM_PRED = "110",
B_PRED = "111" */
};
const tree_index kf_ymode_tree [2 * (num_ymodes - 1)] =
{
-B_PRED, 2, /* root: B_PRED = "0", "1" subtree */
4, 6, /* "1" subtree has 2 descendant subtrees */
-DC_PRED, -V_PRED, /* "10" subtree: DC_PRED = "100",
V_PRED = "101" */
-H_PRED, -TM_PRED /* "11" subtree: H_PRED = "110",
TM_PRED = "111" */
};
const tree_index uv_mode_tree [2 * (num_uv_modes - 1)] =
{
-DC_PRED, 2, /* root: DC_PRED = "0", "1" subtree */
-V_PRED, 4, /* "1" subtree: V_PRED = "10",
"11" subtree */
-H_PRED, -TM_PRED /* "11" subtree: H_PRED = "110",
TM_PRED = "111" */
};
/* Given a bool_decoder d, a Y mode might be decoded as follows.*/
const Prob pretend_its_huffman [num_ymodes - 1] =
{ 128, 128, 128, 128};
Ymode = (intra_mbmode) treed_read( d, ymode_tree,
pretend_its_huffman);
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
{:lang="c"}
Since it greatly facilitates re-use of reference code, and since there is no real reason to do otherwise, it is strongly suggested that any decoder implementation use exactly the same enumeration values and probability table layouts as those described in this document (and in the reference code) for all tree-coded data in VP8.