Commit Graph

384 Commits

Author SHA1 Message Date
Marco
a340c64a79 vp9: Fix some issues with denoiser and SVC.
Fix the update of the denoiser buffer when the base
spatial layer is a key frame. And allow for better/lower
QP on high spatial layers when their base layer is key frame.

Change-Id: I96b2426f1eaa43b8b8d4c31a68b0c6d68c3024a2
2017-03-15 17:19:17 -07:00
Marco
7f2daa74a0 vp9: Incorporate source sum_diff into non-rd partition thresholds.
Increase the variance partition thresholds for superblocks that
have low sum-diff (from source analysis prior to encoding frame).
Use it for now only for speed >= 7 or for denoising on.

Small change on metrics for rtc set: less than ~0.1 avgPNSR decrease
on RTC set, for both speed 7 and 8.

Change-Id: I38325046ebd5f371f51d6e91233d68ff73561af1
2017-02-21 17:22:11 -08:00
paulwilkins
d218b0914e cosmetics: Fix spelling mistake in compile flag name.
agressive -> aggressive

after:
ce7b38459 Aggressive VBR method.

Change-Id: Ie0f30b1bbc77ed9f32bec047b4a9b3d0cf4853f5
2017-02-16 14:51:31 -08:00
Marco
f82280820a vp9. Use same source_sad threshold for all speeds.
Only affects real-time mode.

Change-Id: Iba836f110c4da936f5173cc0f54424d5b6121bff
2017-02-15 11:28:26 -08:00
Paul Wilkins
7635ee0f37 Merge "Aggressive VBR method." 2017-02-15 10:37:02 +00:00
James Zern
3c4ea94210 cosmetics,vp9_ratectrl: apply clang-format
broken since:
c3f095c8b Merge "Fix to avoid abrupt relaxation of max qindex in recode path"
5f21aba4b Fix to avoid abrupt relaxation of max qindex in recode path

the original change pre-dated the addition of .clang-format

Change-Id: If5e399d9a805bcad9147360b13b36fbc8c560a7c
2017-02-13 11:29:39 -08:00
paulwilkins
ce7b38459a Aggressive VBR method.
VBR method that allows a wider Q range for the first normal frame
in each ARF group and then centers the min - max range for the rest of
the arf group on the chosen Q value for that first frame.

This allows for quite rapid adjustment of the active Q range even if the
initial estimate is poor.

In some cases where the ARF frames themselves are tending to
undershoot but the normal frames are overshooting this can still give
net undershoot. This can be corrected by allowing a larger Q delta for
arf frames but is usually is a sign that the allocation to the arfs was to
high.

Change-Id: Icec87758925d8f7aeb2dca29aac0ff9496237469
2017-02-13 15:42:11 +00:00
Paul Wilkins
c3f095c8b3 Merge "Fix to avoid abrupt relaxation of max qindex in recode path" 2017-02-09 17:17:55 +00:00
Marco
3c2f076ad0 vp9: Adjust rate_err threshold for setting active_worst factor.
Only affects 1 pass vbr.
Small improvement on ytlive set.

Change-Id: I09a7456fe658fbea82ece1035cf683bd8bd8bd14
2017-02-07 09:38:16 -08:00
Marco
db99840bf6 vp9: Fixes for usage of skin_map for high bit depth.
Also avoid noise_estimation and source_sad if use_highbitdepth is set.

Change-Id: I5fea396b8f8380ea377045d99ba22a52b92daa46
2017-01-26 19:57:59 -08:00
Jerome Jiang
40ffa2839f vp9: Adjust the threshold to set avg_source_sad_sb flag.
Affect only speed 8. Small/Negligible regression on rtc set.

Change-Id: I67a6b6b4008a22ed798bd980336d95bb799f64b4
2017-01-23 16:11:28 -08:00
Marco
219cdab676 vp9: Add feature to use block source_sad for realtime mode.
Only for speed >= 7, and affects skipping of intra modes.
Threshold is set low for now, needs to be tuned.
Small/no difference in metrics on rtc clips.

Change-Id: If9bdbd43f08d1f80407cdd2e9e5e96780dcd2424
2017-01-20 11:57:02 -08:00
Ranjit Kumar Tulabandu
5f21aba4b0 Fix to avoid abrupt relaxation of max qindex in recode path
The fix relaxes the max qindex based on the data from previous loop of
coding if output frame size is greater than maximum frame size allowed

Change-Id: Iac1f63ec67559d68766e090a7cbb80b812b2560f
2017-01-16 18:03:27 +05:30
Marco
159cc3b33c vp9: Add speed feature flag for computing average source sad.
If enabled will compute source_sad for every superblock on every frame,
prior to encoding. Off by default, only on for speed=8 when
copy_partition is set.

Change-Id: Iab7903180a23dad369135e8234b7f896f20e1231
2017-01-13 11:52:12 -08:00
Marco
91fc730d83 vp9: 1 pass cbr: Adjustments to usage of gf_cbr_boost and aq=3 mode.
When aq=3 mode is on and the gf_cbr_boost is set: make sure golden frame
is always refreshed, and don't incorporate segement cost in qp setting
on the boosted golden frame.

Better performance on RTC set with gf_cbr_boost on,
for example with gf_cbr_boost=50, gains from ~0.5-3%.

Change-Id: Ie811f5e4d444ff3320bd6e2c1745b2c4c09a8460
2017-01-10 09:42:06 -08:00
Marco
bea22782e9 vp9: 1 pass cbr: Fix to qp clamping when gf_cbr_boost_pct is used.
Avoid the qp-clamping on gf/alt frame if gf_cbr_boost_pct is set.

Change only affect CBR mode when  gf_cbr_boost_pct is set.

Change-Id: I0655ed4f2b047c8ed1ed33a070c17960ad776704
2017-01-09 12:52:50 -08:00
Marco Paniconi
ebe0b57c91 Merge "vp9: 1 pass cbr mode: increase threshold for gf_cbr_boost_pct usage." 2017-01-09 17:23:12 +00:00
Marco
f1909d26f8 vp9: 1 pass cbr mode: increase threshold for gf_cbr_boost_pct usage.
Increase the boost threshold below which GOLDEN update will use same
rate correction factor as INTER_NORMAL.

Improves performance when gf_cbr_boost_pct is set (between 0 and 100)
in CBR mode.

Change-Id: I9f54cc18664786a100b13a416b7137ae03bd0cab
2017-01-06 15:37:10 -08:00
Jerome Jiang
afc8c4836f vp9: Compute source sad for every superblock when partition copy is on.
The source sad could be used to copy the partition without going into
choose_partitioning function to speed up vp9 encoding. Computing source
sad takes little time. Speed test on Android and Linux shows little
encoding time gain (less than 1.4%).

Turned off for now since partition copy is turned off.

Change-Id: I61c9d5b8f22329760cb29a4ee30a7f9c232ce8d3
2017-01-06 17:59:02 +00:00
Marco
5de798f2b2 vp9: Fix to usage of flag USE_ALTREF_FOR_ONE_PASS
The flag USE_ALTREF_FOR_ONE_PASS allows for alt-ref lookahead
in 1 pass vbr (from https://chromium-review.googlesource.com/#/c/365498).
This change is to make sure this macro flag only has effect if
the config flag cpi->oxcf.enable_auto_altef is also on.

No change in ytlive encoding, as USE_ALTREF_FOR_ONE_PASS is not
yet enabled.

Change-Id: I1a69681e4a15c5244581a3dab4587fca08f02e0f
2016-12-14 15:07:38 -08:00
Marco
57c6bf291e 1 pass vbr: Allow for lookahead alt-ref in real-time mode.
For 1 pass vbr real-time mode:
Allow for the usage of alt-ref frame when non-zero lag-in-frames is used.
Use non-filtered alt-ref, and select usage based on fast scene/content
analysis/detection within the lag of frames.

Positive gains on ytlive set: overall avgPSNR ~3-4%.
Several clips are up between 5-14%, a few clips are neutral/small change.

Current speed decrease is about ~5-10%.

Use the flag USE_ALTREF_FOR_ONE_PASS to enable this feature
(off by default for now).

Change-Id: I802d2bf3d44f9cf01f6d15c76be9c90192314769
2016-10-11 10:13:17 -07:00
Marco
cdbd89197e vp9: 1 pass vbr: some adjustments to gf interval.
Put limit on gf interval based on lag, and allow
for the adjustment on next gf group also on key frame.

Small/neutral change on ytlive metrics.

Change only affects 1 pass vbr real-time mode.

Change-Id: I339c8f4398848698b6e10fe9482c52ca661b94a5
2016-10-11 08:34:12 -07:00
James Zern
7a9e476072 Merge changes from topic 'clang-format'
* changes:
  apply clang-format
  .clang-format: update to 3.8.1
2016-09-16 07:11:33 +00:00
Marco
4c1a9fb8db vp9: Small code cleanup.
Remove the experiment LIMIT_QP_ONEPASS_VBR_LAG, as its
not currently used and no plan to use in near future.

Change-Id: Ib069f8d7225195be04b765d0ab477510dfba6a3b
2016-09-15 15:17:17 -07:00
clang-format
5f6d143b41 apply clang-format
Change-Id: I501597b7c1e0f0c7ae2aea3ee8073f0a641b3487
2016-09-15 15:07:53 -07:00
Yaowu Xu
454139ae13 Merge "Casts to remove some warnings." 2016-09-01 23:37:04 +00:00
paulwilkins
3e9e77008c Casts to remove some warnings.
Added casts to remove warnings:
BUG=webm:1274

In regards to the safety of these casts they are of two types:-

- Normalized bits per (16x16) MB stored in a 32 bit int (This is safe as bits
per MB even with << 9 normalization cant overflow 32 bits. Even raw 12
bits hdr source even would only be  29 bits :- (4+4+12+9) and the encoder
imposes much stricter limits than this on max bit rate.

- Cast as part of variance calculations.  There is an internal cast up to 64 bit
for the Sum X Sum calculation, but after normalization dividing by the number
of points the result will always be <= the SSE value.

Change-Id: I4e700236ed83d6b2b1955e92e84c3b1978b9eaa0
2016-09-01 16:10:12 +01:00
paulwilkins
6fc07a217d Modified resize loop constraints.
Using a tighter resize constraint on undershoot seems to help
results (especially SSIM) as significant undershoot on a frame
seems to have more of a damaging impact than overshoot.

This patch has been tuned so that in local testing using the
derf set it is encode speed neutral for speed  setting 2.

Average quality result for speed 2 (psnr,ssim) were  as follows:-

 lowres  0.039,  0.453
 midres  0.249, 0.853
 hdres  0.159, 0.659
 NetFlix -0.241, 0.360

Change-Id: Ie8d3a0d7d6f7ea89d9965d1821be17f8bda85062
2016-08-31 12:45:49 +01:00
clang-format
3a4002b94d vp9_ratectrl.c: apply clang-format
after:
ff0a87c vp9 1pass vbr: Adjustment to gf interval.

Change-Id: I1296e53e601bf0c2b562e3a34082ac45c294a5f1
2016-08-04 11:57:00 -07:00
Marco Paniconi
9fdeeaf411 Merge "vp9 1pass vbr: Adjustment to gf interval." 2016-08-04 17:50:55 +00:00
Marco
ff0a87ce38 vp9 1pass vbr: Adjustment to gf interval.
Increase the minimum distance.
Reduces the overshoot somewhat on some clips,
small gain in avgPSNR (~0.1%) on ytlive set.

Change-Id: Id5ddde20c2907dbdb536e79542eff775019c142b
2016-08-03 15:36:27 -07:00
clang-format
e0cc52db3f vp9/encoder: apply clang-format
Change-Id: I45d9fb4013f50766b24363a86365e8063e8954c2
2016-08-02 16:47:11 -07:00
Marco
c06a4b9df2 vp9 1pass-vbr: Adjust gf setting for nonzero-lag case.
Change-Id: I230c586c6d5ae56ee9a6d37b7d9452351bb4bd80
2016-07-22 11:48:09 -07:00
jackychen
71f9cbcfc8 vp9: Fix the clang warning of unsigned int type.
Change-Id: I6308db16bd626fa5943925471e9171f567669350
2016-07-20 15:58:35 -07:00
Marco
05fe0f20a6 vp9: Allow usage of lookahead for real-time, 1 pass vbr.
Allow usage of lookahead for VBR in real-time mode, for 1 pass vbr.

Current usage is for fast checking of future scene cuts/changes,
and adjusting rate control (gf interval and active_worst/target size).

Added unittests (datarate) for 1 pass vbr mode, with non-zero lag.

Added an experimental option to limit QP based on lookahead.

Overall positive gain in metrics on ytlive set:
avgPNSR/SSIM up on average ~1-3%; several clips up by 5, 7%.

Change-Id: I960d57dfc89de121c4824b9a9bf88d2814e74b56
2016-07-18 15:20:17 -07:00
Marco
cc431ad50a vp9: Adjustment of gfu_boost and af_ratio for 1 pass vbr.
Modify the gfu_boost and af_ratio setting based on the
average frame motion level.

Change only affects 1 pass vbr.

Metrics overall positive on ytlive set.
On average up by ~1%, several clips up by 2-4%.

Change-Id: Ic18c49eb2df74cb4986b63cdb11be36d86ab5e8d
2016-07-07 15:18:14 -07:00
Marco
9b61c709c4 vp9: Put lower limit on frames_to_key adjustment to gf interval.
For 1 pass vbr.

Change-Id: I6c1c1cca3944150140e896e175b07ed518641b9b
2016-06-10 10:08:56 -07:00
Marco
77ffea92c5 vp9: Reduce overshoot factor for setting longer gf interval.
Only affects 1 pass vbr mode.

Change-Id: Ifd7ff3771eabe322a371b35b159cace6baa3b070
2016-06-08 15:15:15 -07:00
Marco
f0b08aa4fa vp9: Replace the avg_size_inter metric, for 1 pass vbr.
Code cleaup, use existing rolling_actual/target metrics instead,
set threshold to get same/similar effect.

Little/no change in metrics on ytlive set.

Change-Id: I74f3c3d0a143a9cf20dc9c3dee54c0f7e6a97a51
2016-06-07 11:29:33 -07:00
Marco
3de42ac27b vp9: Small ajustment to settings gf_interval, 1 pass vbr.
Add a max condition and lower the min value.

No change in behavior (metrics for yt live set) for the
default min/max_gf_interval=4/16 settings.
Small positive change when min/max_gf_interval=7/16
(for 60fps clips on ytlive set).

Change-Id: I1c1d72425c86c69419ea43fb9730130e81062f91
2016-06-07 09:51:41 -07:00
paulwilkins
c7ac2f3864 Adjustment to VBR rate correction.
Changes to the function the redistributes bits from overshoot
or undershoot throughout the rest of the clip to respond more
quickly.

Change-Id: I90f10900cdd82cf2ce1d8da4b6f91eb5934310da
2016-06-03 13:17:43 +01:00
paulwilkins
4328b08521 Remove gf_zeromotion_pct.
The use of this value is preventing rate adjustment on clips
or sections that have very little motion but high noise and
this can give rise to some sections with massive overshoot.

Change-Id: I9a65c7c1148dc5d3a7d8b23e50fc1733f3661621
2016-06-03 12:13:03 +01:00
Marco
8a79769aff vp9: Update to rc-metric for keeping track of average frame size.
Rename and change to how its updated.
Only affects 1 pass vbr.

Small change in metrics (< ~0.1%) on ytlive set.

Change-Id: Ibb1fe485699b6c4a8194951c8f229abe2f64b9a5
2016-05-13 14:31:57 -07:00
Marco
0c60db7def vp9: Fix to quality regression issue for 1 pass.
BUG=https://bugs.chromium.org/p/webm/issues/detail?id=1217

Issue introduced in https://chromium-review.googlesource.com/#/c/339162/.
The factor on avg_frame_qindex[INTER_FRAME] in that CL, used for q-basis to
set active_best, can cause the QP to decrease too slowly or get stuck
in some cases when max-q=63.

Removing that factor fixes the regression in issue#1217.

Change-Id: I97fe28dca3c2bc2d852b1775f39c62164e032d4f
2016-05-11 14:01:21 -07:00
Marco
36f796d2d0 vp9: 1 pass vbr: Reduce base qp for active_best_quality for inter-frames.
Reduce factor for setting base-qp for active_best_quality (for inter-frames).
Small increase in metrics on yt live set.

Change-Id: I9cf0ac797783aeddbfaf1ff510696c9035d7c5ee
2016-05-09 10:21:44 -07:00
Marco
1fc21052ab vp9: Modify logic for gf setting based on up-coming key frame.
For 1 pass vbr mode:
Refactor to move the logic for gf setting based on up-coming
key frames to a separate function, so same logic can be used for
scene-cuts/changes.

Change-Id: Ic4ede308e08ba869bb62e4566e19ea31222c5229
2016-05-05 09:33:27 -07:00
Marco
7bd73b638a vp9: Move the initialization of some rc variables to rc_init.
And initialize a variable (avg_source_sad) that was not initialized to 0.

Change-Id: I63223e20fffe2a2ea38f03da62dad16af0618ea4
2016-04-28 17:10:37 -07:00
Marco
e172e8aa26 vp9: 1 pass vbr: Adjust gf boost/interval on motion level.
Add some adjustment for high motion case.

Change-Id: If700be1ce849c187ed031221ea7cabbc8d91ec2b
2016-04-25 11:54:10 -07:00
Marco
8b25dd599f vp9: Add rc quantity to track amount of low motion in scene.
Use it for now in noise estimation to bypass estimation if
motion level is high.

Change-Id: I033662dc909f2060e4e81abf562a7ad262dc8170
2016-04-25 09:07:35 -07:00
Marco
83f17eeede vp9: 1 pass vbr: Small fix on reset avg_intersize_gfint on scene cut.
Change-Id: Ie73d1bbc2c953d010b32097440698f0d060750be
2016-04-20 16:32:09 -07:00