diff options
| author | Ronald S. Bultje <rsbultje@gmail.com> | 2015-10-13 14:06:28 -0400 |
|---|---|---|
| committer | Ronald S. Bultje <rsbultje@gmail.com> | 2015-10-16 19:30:38 -0400 |
| commit | 6e5a1165bef829b3fcc73c36e5c950592b5c6801 (patch) | |
| tree | 477f0a118cebdbf8d00d30070245bb405b5dcc58 /vp10/encoder/encoder.h | |
| parent | 7dd7a7da206bddb7d3d3deffa23984719bb89d87 (diff) | |
| download | libvpx-6e5a1165bef829b3fcc73c36e5c950592b5c6801.tar libvpx-6e5a1165bef829b3fcc73c36e5c950592b5c6801.tar.gz libvpx-6e5a1165bef829b3fcc73c36e5c950592b5c6801.tar.bz2 libvpx-6e5a1165bef829b3fcc73c36e5c950592b5c6801.zip | |
vp10: make segmentation probs use generic probability model.
Locate them (code-wise) in frame_context, and have them be updated
as any other probability using the subexp forward and adaptive bw
updates.
See issue 1040 point 1.
TODOs:
- real-world default probabilities
- why is counts sometimes NULL in the decoder? Does that mean bw
adaptivity updates only work on some frames? (I haven't looked
very closely yet, maybe this is a red herring.)
Change-Id: I23b57b4e5e7574b75f16eb64823b29c22fbab42e
Diffstat (limited to 'vp10/encoder/encoder.h')
| -rw-r--r-- | vp10/encoder/encoder.h | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/vp10/encoder/encoder.h b/vp10/encoder/encoder.h index a957b6382..98ded30d1 100644 --- a/vp10/encoder/encoder.h +++ b/vp10/encoder/encoder.h @@ -55,7 +55,9 @@ typedef struct { int nmvcosts[2][MV_VALS]; int nmvcosts_hp[2][MV_VALS]; +#if !CONFIG_MISC_FIXES vpx_prob segment_pred_probs[PREDICTION_PROBS]; +#endif unsigned char *last_frame_seg_map_copy; |