Age | Commit message (Collapse) | Author |
|
|
|
Change-Id: Ifc165d76a71fcdb7c19c158c940a8d273be0d95f
|
|
BUG=webm:1591
Change-Id: I926566ac1bf4bac8cb1ce1c6ded9ba940109283e
|
|
this matches what is done to reduce the cost of the test of filter
values in convolve.
Change-Id: I692b58801a962b593b810c1d1dac42f72c78caf9
|
|
|
|
|
|
For screen-content mode, with aq-mode=3: increase the
qp thresh for disabling the cyclic refresh.
Improves bitrate convergence for content that has been
static for long period.
Change-Id: Ica63a741402923a611ab1b86c0900f75d2d5f941
|
|
* changes:
Implement get_mv_cost()
Add assertion in get_block_src_pred_buf
Fix bug in predict_mv_mode
Allocate memory for mv_mode_arr[]
|
|
For non-rd pickmode: include H and V intra mode check for
spatially flat blocks when the sf->short_circuit_flat_blocks
speed feature is set.
Small improvement on screen content tests.
Change-Id: I3391d02cce6a46160be6ccc8a1e33fd8547eb467
|
|
The mv_cost contains mv_mode cost and mv_diff cost.
The mv_mode cost is inferred from default_inter_mode_probs.
The mv_diff cost is estimated used the log2 function.
Change-Id: I62702bdb5c3fec018e3302765f5dd749fceebc12
|
|
This changes the highbd version of temporal filter to information from
both luma and chroma planes.
Performance:
AVG_PSNR | OVR_PSNR | SSIM
-0.144% | -0.165% | -0.150%
The performance is evaluated on lowres_bd10.
Change-Id: I89d1bd46cd60c26d658b6a53aa63835e90d8e291
|
|
For screen content mode: always force intra check
for spatially flat blocks that have moved. Also
adjust/fix condition for forcing check of
zeromv-golden for quality layers.
Reduces artifacts in screensharing tests.
Change-Id: Iafd62fb24a4e05f5b12af663dde2805fdb4c7b36
|
|
Print error message and assert when ref_frame_idx is invalid
Change-Id: Ice6d06c53ddae0a77d578671b896c4e4d04d5366
|
|
* changes:
Modify map read/write to sync logic in row_mt case
Revert "Revert "Add Tile-SB-Row based Multi-threading in Decoder""
|
|
|
|
|
|
|
|
Modify early breakout condition for non-rd pickmode
for quality layers: when lower layer has lower QP force
test of zeromv on golden (lower layer reference) before
breakout due to skip.
Reduce artifacts, observed in cases of scrolling content.
Change-Id: Id834b1eb024a4c97f0e74d8b7f7a0351459e088f
|
|
Adds conditional wait/signal instead of sched_yield.
Change-Id: I49a760eacdd6b6ac690e797ea5f10febf6a1a084
|
|
Use kMvPreCheckLines in the for loops.
Change-Id: I60204c6294560d47421a8621d907dfa95c9dde18
|
|
Change-Id: Ib9fdf65e263dbaace8d4c86766eba2c6f35f652b
|
|
Change-Id: Ib820f625e0b616fd57a2722ec3614b4fccf307f8
|
|
In the ML based partition search speed feature, use MV result of
previous simple motion search as the starting point for the next one.
Compression change is neutral; encoding speed becomes slightly faster.
Change-Id: Iea554f28f7966fc5b5857e12b06de58e3fa312a6
|
|
* changes:
Add predict_mv_mode_arr()
Add predict_mv_mode()
|
|
This reverts commit a4d2f59b691ac57698d299430495b398a8bb682c.
Reason for revert: Re-enables SSE4_1 version of apply temporal filter now that the mismatch is fixed in fa540837aa23ae8cdd926961ddbef71998a2a525,
Original change's description:
> Revert "Enable SSE4 version of apply temporal filter"
>
> This reverts commit 4f3cd48bfe4541ca64883db51f57c1376c6d8a73.
>
> Reason for revert: Found a mismatch with c version
>
> Original change's description:
> > Enable SSE4 version of apply temporal filter
> >
> > Evaluating on 5 midres clips with 4 bitrates over 30 frames on speed 1
> > auto_alt_ref=1, there is a speed up of 1.660%.
> >
> > BUG=webm:1591
> >
> > Change-Id: Idbda58548679e6f7b8fc0d7f6144f7be057ef690
>
> TBR=yunqingwang@google.com,builds@webmproject.org,chiyotsai@google.com
>
> Change-Id: Ibca973576d72d6db4b647a08aef23389d5d6605a
> No-Presubmit: true
> No-Tree-Checks: true
> No-Try: true
> Bug: webm:1591
TBR=yunqingwang@google.com,builds@webmproject.org,chiyotsai@google.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Bug: webm:1591
Change-Id: I26effdbaf4d52e4650c263b6ed9d3d80e505f5cb
|
|
|
|
Change-Id: I6503ebc79beaac2947992437ac133f3ac4379019
|
|
Refactor out code about simple motion search.
Change-Id: Ie6895db2aff3c13e7a45554d6bc1c7c0af8f2d51
|
|
The function predicts the mv_mode for each prediction block in
diagonal order.
Change-Id: I7dcfcdb317ffa334cb40bb435baa71b5db62252b
|
|
This function evaluate the impact of setting NEW_MV_MODE on a
block and its neighbor blocks.
Change-Id: Ie0b2c67bdc5cd14e0efd8ebc5dc3f3f873bcf3fe
|
|
|
|
This reverts commit 4f3cd48bfe4541ca64883db51f57c1376c6d8a73.
Reason for revert: Found a mismatch with c version
Original change's description:
> Enable SSE4 version of apply temporal filter
>
> Evaluating on 5 midres clips with 4 bitrates over 30 frames on speed 1
> auto_alt_ref=1, there is a speed up of 1.660%.
>
> BUG=webm:1591
>
> Change-Id: Idbda58548679e6f7b8fc0d7f6144f7be057ef690
TBR=yunqingwang@google.com,builds@webmproject.org,chiyotsai@google.com
Change-Id: Ibca973576d72d6db4b647a08aef23389d5d6605a
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: webm:1591
|
|
|
|
* changes:
Add find_best_ref_mv_mode()
Add get_mv_dist
Add get_mv_from_mv_mode()
|
|
|
|
|
|
|
|
Adjustments to the calculation and use of a noise estimate in
the first pass Q estimate and adaptation of temporal filtering.
This change was tested and gave gains for both auto-alt-ref=1
and auto-alt-ref=6 as follows:
Results are Av PSNR, Overall PSNR, SSIM and PSNR-HVS
auto-alt-ref=1
low_res 0.007, -0.042, -0.018, 0.074
mid_res -0.142, -0.239, -0.173, -0.129
hd_res -0.322, -0.405, -0.397, -0.367
NF_2K -0.058, -0.099, -0.201, 0.028
auto-alt-ref=6
low_res -0.058, -0.171, -0.188, -0.027
mide_res -0.149, -0.155, -0.171, -0.137
hd_res -0.252, -0.339, -0.259, -0.297
NF_2K -0.015, -0.068, -0.120, 0.092
In all sets there were some winners and losers but significantly
more winners. The biggest change was Stockholm in the
hd set with an improvement of 5-6%
Change-Id: Ieec71e1c4e3e09b76c288efa7b4d1b00015b3a11
|
|
Evaluating on 5 midres clips with 4 bitrates over 30 frames on speed 1
auto_alt_ref=1, there is a speed up of 1.660%.
BUG=webm:1591
Change-Id: Idbda58548679e6f7b8fc0d7f6144f7be057ef690
|
|
This adds a preliminary version of vp9_apply_temporal_filter in SSE4.1.
This patch merely adds the function and does not enable it yet.
Speed Up:
| ss_x=1 | ss_x=0 |
ss_y=1 | 19.80X | 19.04X |
ss_y=0 | 21.09X | 20.21X |
BUG=webm:1591
Change-Id: If590f1ccf1d0c6c3b47410541d54f2ce37d8305b
|
|
This function compute the rd cost for each mv_mode and return the
one with minimum rd cost.
eval_mv_mode()
Evaluate the rd cost for a given mv_mode.
Change-Id: Ia1b3ec7e1dd538e443e1bc79f2cab352408cd0a0
|
|
Given an mv_mode, get_mv_dist() obtains the mv and uses it
to compute distortion.
Change-Id: I58b8c7137b99c2736d651e678f0cd013dbd94877
|
|
* changes:
Add set_block_src_pred_buf()
[cleanup] Move get_feature_score to a proper place
|
|
Given an mv_mode, this function will return the corresponding mv.
find_ref_mv()
A helper function finds the nearest and near mvs from the neighbor
blocks.
select_mv_arr[]
An array used for storing selected motion vectors.
Change-Id: Ibeb434007f65b2c6e461360f208d99455e76bcbf
|
|
Avoid recursively decreasing 'strength'.
avg_psnr ovr_psnr ssim
midres -0.224 -0.195 -0.115
Change-Id: Ie74c069cda76873ac38e9c1a9162b1ddfb9b103d
|
|
This function sets src and pre buffer of MACROBLOCK
and MACROBLOCKD.
Will add static decorator once this function is called.
Change-Id: I0fb46784dd97839e4d87c9e027fe8c59683e70d8
|
|
Add static decorator to it as well.
Change-Id: I6c89fae456561b6975ab49af139a45a7483507c6
|
|
Exclude low bit depth optimizations from high bit depth builds.
BUG=webm:1584
Change-Id: I86a7ebafa557d262257358e1e055a06d52659977
|
|
This reverts commit 06983668cf41f66765528db044419f954e5a5d64.
Fixes Visual Studio build errors introduced by earlier row mt commit
BUG=webm:1587
Change-Id: I792df86e8254cd6b2a511955b691af619a569cd0
|
|
This commit introduces a new speed feature that determines the
SEARCH_METHOD used by temporal filter when doing 16x16 block on
full_pixel_motion_search. On speed 0, the most exhaustive method MESH is
used. On speed 1 and above, a faster method NSTEP is used.
Performance:
| AVG_PSNR | AVG_SPDUP | AVG_SPDUP:AVG_PSNR
MISRES | 0.007% | 2.818% | 402:1
HDRES | 0.004% | 4.897% | 1224:1
In the case of midres, there is a small quality gain of -0.021% on
OVR_PSNR.
Performance measurement is done on speed 1 with auto_alt_ref=1.
Quality is measured on full midres set over 60 frames. Speed is measured
on 5 midres clips over 4 bitrates over 30 frames.
STATS_CHANGED
Change-Id: Ic1879d2237f8734529e194767a6cf5e43e20b47b
|