diff options
| author | Deb Mukherjee <debargha@google.com> | 2012-09-06 09:07:42 -0700 |
|---|---|---|
| committer | Deb Mukherjee <debargha@google.com> | 2012-09-12 11:51:10 -0700 |
| commit | 2b26cf17862949123618d2dbe60a661fbb7eb4a9 (patch) | |
| tree | ef2cbadc62a86cad8c5d69a7287159367ef397f7 /vp8/decoder | |
| parent | d406334f2712a1ec1bef890b2397927918a84e48 (diff) | |
| download | libvpx-2b26cf17862949123618d2dbe60a661fbb7eb4a9.tar libvpx-2b26cf17862949123618d2dbe60a661fbb7eb4a9.tar.gz libvpx-2b26cf17862949123618d2dbe60a661fbb7eb4a9.tar.bz2 libvpx-2b26cf17862949123618d2dbe60a661fbb7eb4a9.zip | |
Adds feature for companded MV encoding
The high-precision (1/8) pel bit is turned off if the reference
MV is larger than a threshold. The motivation for this patch is
the intuition that if motion is likely large (as indicated by
the reference), there is likley to be more motion blur, and as
a result 1/8 pel precision would be wasteful both in rd sense
as well as computationally.
The feature is incorporated as part of the newmventropy experiment.
There is a modest RD improvement with the patch. Overall the
results with the newmventropy experiment with the threshold being
16 integer pels are:
derf: +0.279%
std-hd: +0.617%
hd: +1.299%
yt: +0.822%
With threshold 8 integer pels are:
derf: +0.295%
std-hd: +0.623%
hd: +1.365%
yt: +0.847%
Patch: rebased
Patch: rebase fixes
Change-Id: I4ed14600df3c457944e6541ed407cb6e91fe428b
Diffstat (limited to 'vp8/decoder')
| -rw-r--r-- | vp8/decoder/decodemv.c | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/vp8/decoder/decodemv.c b/vp8/decoder/decodemv.c index 8d0f94e65..7f7567e4e 100644 --- a/vp8/decoder/decodemv.c +++ b/vp8/decoder/decodemv.c @@ -244,6 +244,7 @@ static void read_nmv(vp8_reader *r, MV *mv, const MV *ref, static void read_nmv_fp(vp8_reader *r, MV *mv, const MV *ref, const nmv_context *mvctx, int usehp) { MV_JOINT_TYPE j = vp8_get_mv_joint(*mv); + usehp = usehp && vp8_use_nmv_hp(ref); if (j == MV_JOINT_HZVNZ || j == MV_JOINT_HNZVNZ) { mv->row = read_nmv_component_fp(r, mv->row, ref->row, &mvctx->comps[0], usehp); @@ -252,6 +253,7 @@ static void read_nmv_fp(vp8_reader *r, MV *mv, const MV *ref, mv->col = read_nmv_component_fp(r, mv->col, ref->col, &mvctx->comps[1], usehp); } + //printf(" %d: %d %d ref: %d %d\n", usehp, mv->row, mv-> col, ref->row, ref->col); } static void update_nmv(vp8_reader *bc, vp8_prob *const p, |