platform/upstream/libvpx.git
5 years agoAllow macroblock_plane to have its own rounding buffer
Jingning Han [Mon, 25 Mar 2019 22:35:35 +0000 (15:35 -0700)]
Allow macroblock_plane to have its own rounding buffer

Add 8 bytes buffer to macroblock_plane to support rounding factor.

Change-Id: I3751689e4449c0caea28d3acf6cd17d7f39508ed

5 years agotool/set_analyzer_env: disable implicit-integer-truncation
James Zern [Mon, 1 Apr 2019 19:03:22 +0000 (12:03 -0700)]
tool/set_analyzer_env: disable implicit-integer-truncation

with clang-7 this causes additional warnings in x86 intrinsics and
elsewhere. disabling for now to unblock new changes.

BUG=webm:1615

Change-Id: Ide9cacee5547ed432f980f6804e1414f32639121

5 years agoMerge "Use reconstructed variance in rd_variance_adjustment"
Paul Wilkins [Mon, 1 Apr 2019 11:57:13 +0000 (11:57 +0000)]
Merge "Use reconstructed variance in rd_variance_adjustment"

5 years agoMerge "update .clang-format for version clang-7.0.1 update."
James Zern [Fri, 29 Mar 2019 18:59:40 +0000 (18:59 +0000)]
Merge "update .clang-format for version clang-7.0.1 update."

5 years agoupdate .clang-format for version clang-7.0.1 update.
Hien Ho [Fri, 29 Mar 2019 16:04:41 +0000 (09:04 -0700)]
update .clang-format for version clang-7.0.1 update.

added files that are affected by clang-format version 7.

BUG=b/120815481

Change-Id: I40662ce962e4f4b1fcdf183b700f85cc5c0f9f82

5 years agoMerge "Revert "Wrap macro definition in do-while(0)""
Jerome Jiang [Thu, 28 Mar 2019 18:00:57 +0000 (18:00 +0000)]
Merge "Revert "Wrap macro definition in do-while(0)""

5 years agoRevert "Wrap macro definition in do-while(0)"
Jerome Jiang [Thu, 28 Mar 2019 17:03:27 +0000 (10:03 -0700)]
Revert "Wrap macro definition in do-while(0)"

This reverts commit aa04b6f9a7475e9d9457dfc5bf441faf15efc466.

It caused big regression on webrtc VP8 tests.

Change-Id: I937e769d133abeca62ba063e59a58b5c461f5b5e

5 years agoMerge "Disable mismatch check on vp9 svc examples."
Jerome Jiang [Thu, 28 Mar 2019 16:26:49 +0000 (16:26 +0000)]
Merge "Disable mismatch check on vp9 svc examples."

5 years agoUse reconstructed variance in rd_variance_adjustment
sdeng [Mon, 18 Mar 2019 18:26:06 +0000 (11:26 -0700)]
Use reconstructed variance in rd_variance_adjustment

to replace the variance from .dst which is the prediction buffer in
inter mode.  Only enable it in tune-content-film mode at the moment.

Change-Id: I647b4a524a0849fda42541887ebc34091f152073

5 years agoMerge "Allow more Intra choices in film mode."
Paul Wilkins [Thu, 28 Mar 2019 10:49:55 +0000 (10:49 +0000)]
Merge "Allow more Intra choices in film mode."

5 years agoDisable mismatch check on vp9 svc examples.
Jerome Jiang [Wed, 27 Mar 2019 23:12:04 +0000 (16:12 -0700)]
Disable mismatch check on vp9 svc examples.

Change-Id: I49902a750758ba0ffe733be9b1efd0cdea44f936

5 years agovp9-rtc: Fix noise estimation thresh for 1080p
Marco Paniconi [Wed, 27 Mar 2019 21:19:04 +0000 (14:19 -0700)]
vp9-rtc: Fix noise estimation thresh for 1080p

Use the same thresholds as for 720p for now,
leads to better noise estimation on test clips.

Change-Id: I55e11346a747fe149f521315a38d75e28b3e774e

5 years agoAllow more Intra choices in film mode.
Paul Wilkins [Wed, 27 Mar 2019 17:57:00 +0000 (17:57 +0000)]
Allow more Intra choices in film mode.

Disable part of a speed feature that blocks all intra modes
except DC_PRED when the source variance is low.

Change-Id: I2956951fd05933a39f7225d4dfe14e019410fee3

5 years agovp9-rtc: Disable cyclic refresh under some conditions
Marco Paniconi [Tue, 26 Mar 2019 17:45:22 +0000 (10:45 -0700)]
vp9-rtc: Disable cyclic refresh under some conditions

cyclic refresh does not work for speeds <= 4, so disable
it for this case. And dynamically disable it when
average_qp is close to MAXQ (only for non-svc), to improve
quality/rate control at very low bitrates.

Change-Id: I447be43aef0fbb80f4a30d81e11658b58744eae5

5 years agoMerge "Remove deprecated code for vp9_fdct8x8_quant()"
Jingning Han [Mon, 25 Mar 2019 23:50:01 +0000 (23:50 +0000)]
Merge "Remove deprecated code for vp9_fdct8x8_quant()"

5 years agoMerge "Unify the transform and quantization process"
Jingning Han [Mon, 25 Mar 2019 23:49:52 +0000 (23:49 +0000)]
Merge "Unify the transform and quantization process"

5 years agoRemove deprecated code for vp9_fdct8x8_quant()
Jingning Han [Mon, 25 Mar 2019 22:02:16 +0000 (15:02 -0700)]
Remove deprecated code for vp9_fdct8x8_quant()

Change-Id: If146bbf24f446f71be9147402e6d30533eee99d1

5 years agoUnify the transform and quantization process
Jingning Han [Mon, 25 Mar 2019 18:27:14 +0000 (11:27 -0700)]
Unify the transform and quantization process

Unify the transform and quantization process for 4x4 - 16x16
transform block sizes. This doesn't affect the encoding speed
visibly. Remove it to reduce the maintenance load.

Change-Id: Ifbf20bf8554ecf7970a6279a2b783b1c58fac6e4

5 years agoChange PSNR threshold for high bitdepth.
Jerome Jiang [Mon, 25 Mar 2019 20:42:31 +0000 (13:42 -0700)]
Change PSNR threshold for high bitdepth.

BUG=webm:1609

Change-Id: I1aa18d58c20532f657059a2df3646fad1625e3ae

5 years agoUse high bd path to substract blocks in hbd build.
Jerome Jiang [Sat, 23 Mar 2019 00:27:24 +0000 (17:27 -0700)]
Use high bd path to substract blocks in hbd build.

BUG=webm:1609

Change-Id: Ifc15d616e7cfb247b399def64ef7691589d90075

5 years agoMerge "Use most common segment index for 64x64 level decision"
Jingning Han [Fri, 22 Mar 2019 05:01:57 +0000 (05:01 +0000)]
Merge "Use most common segment index for 64x64 level decision"

5 years agoMerge "Add wiener variance log function"
Jingning Han [Fri, 22 Mar 2019 05:01:51 +0000 (05:01 +0000)]
Merge "Add wiener variance log function"

5 years agoMerge "Refactor Wiener variance based segmentation"
Jingning Han [Fri, 22 Mar 2019 05:01:43 +0000 (05:01 +0000)]
Merge "Refactor Wiener variance based segmentation"

5 years agofix redundant cast in examples.
Jerome Jiang [Fri, 22 Mar 2019 02:04:38 +0000 (19:04 -0700)]
fix redundant cast in examples.

Change-Id: I84280de82053f9056bda9d813baa6165ca9bcd1e

5 years agoUse most common segment index for 64x64 level decision
Jingning Han [Thu, 21 Mar 2019 22:40:21 +0000 (15:40 -0700)]
Use most common segment index for 64x64 level decision

Find the most common segment index among all 16x16 blocks in a
64x64 block and use that as the 64x64 block level decision.

Change-Id: I67e85869d9fee0fc05450928f1eeaebe511cab6a

5 years agoAdd wiener variance log function
Jingning Han [Thu, 21 Mar 2019 22:15:31 +0000 (15:15 -0700)]
Add wiener variance log function

Factor common mapping function from wiener variance to its log
form.

Change-Id: I25c955c8a3e25b9af1d65a0f0c3f695547c13453

5 years agoMerge "vp9 postproc neon: Remove the condition on mb cols."
Jerome Jiang [Thu, 21 Mar 2019 22:24:20 +0000 (22:24 +0000)]
Merge "vp9 postproc neon: Remove the condition on mb cols."

5 years agoRefactor Wiener variance based segmentation
Jingning Han [Thu, 21 Mar 2019 22:06:13 +0000 (15:06 -0700)]
Refactor Wiener variance based segmentation

Separate the k-means clustering stage and the segmentation parse
stage to save unnecessary steps in a common function.

Change-Id: I60083e1d970e744f9a64112f856892d450f86669

5 years agovp9 postproc neon: Remove the condition on mb cols.
Jerome Jiang [Thu, 21 Mar 2019 21:17:15 +0000 (14:17 -0700)]
vp9 postproc neon: Remove the condition on mb cols.

VP8 and VP9 have different padding on buffer stride.
VP8 microblock is 16x16 so the buffer stride needs to be divisible by
16. Thus UV buffer stride is divisible by 8.
VP9 microblock is 8x8 so the buffer stride is only extended to be
divisible by 8. Then UV buffer stride isn't divisible by 8.

Change-Id: I6fa953feb951f2fb2e48f72a623786b85e23822f

5 years agoMerge "vp9: remove condition on high bitdepth using simple block yrd."
Jerome Jiang [Thu, 21 Mar 2019 19:49:51 +0000 (19:49 +0000)]
Merge "vp9: remove condition on high bitdepth using simple block yrd."

5 years agovp9-screen: Adjust speed features at speed 8
Marco Paniconi [Thu, 21 Mar 2019 17:36:31 +0000 (10:36 -0700)]
vp9-screen: Adjust speed features at speed 8

Keep loopfilter on, and use half-pel instead of full.
This reduces big quality gap between the speed 8 and 7,
but still keeps speed 8 about 30-40% faster than speed 7.
Tested on screenshare clips with scroll and slide changes.

Change-Id: Id63b44f59655f3e3dc1b49d89291d97e7323081a

5 years agoMerge "vp9-rtc: Fixes for low-resoln"
Marco Paniconi [Wed, 20 Mar 2019 23:06:05 +0000 (23:06 +0000)]
Merge "vp9-rtc: Fixes for low-resoln"

5 years agoMerge "Properly reset memory in hbd setting"
Jingning Han [Wed, 20 Mar 2019 21:31:21 +0000 (21:31 +0000)]
Merge "Properly reset memory in hbd setting"

5 years agoMerge "Drop Wiener variance diff scale from 8 to 4"
Jingning Han [Wed, 20 Mar 2019 21:31:12 +0000 (21:31 +0000)]
Merge "Drop Wiener variance diff scale from 8 to 4"

5 years agoMerge "Enable all the 8 segments by default for perceptual AQ mode"
Jingning Han [Wed, 20 Mar 2019 21:31:03 +0000 (21:31 +0000)]
Merge "Enable all the 8 segments by default for perceptual AQ mode"

5 years agovp9-rtc: Fixes for low-resoln
Marco Paniconi [Wed, 20 Mar 2019 20:39:48 +0000 (13:39 -0700)]
vp9-rtc: Fixes for low-resoln

The force smooth_filter should only be used
for noisy content, so for now keep it off and
add TODO. Also fix/adjust low-resoln condition
and threshold in cyclic refresh.

Change-Id: I6c456dc9f23daabba20badd65a2f7ee6c5e259c4

5 years agoProperly reset memory in hbd setting
Jingning Han [Wed, 20 Mar 2019 20:19:01 +0000 (13:19 -0700)]
Properly reset memory in hbd setting

This avoids a segmentation failure issue in high bit-depth case.

Change-Id: I9fbb3ec24b1735678f110cb084a29b15e3ec1a12

5 years agoDrop Wiener variance diff scale from 8 to 4
Jingning Han [Wed, 20 Mar 2019 18:14:57 +0000 (11:14 -0700)]
Drop Wiener variance diff scale from 8 to 4

Change-Id: If6df62069c1b30de4d52aba33c20c6e854524905

5 years agoEnable all the 8 segments by default for perceptual AQ mode
Jingning Han [Wed, 20 Mar 2019 18:13:22 +0000 (11:13 -0700)]
Enable all the 8 segments by default for perceptual AQ mode

Change-Id: I8999ee74216785c22568a09bce7590c9fc6905c1

5 years agoMerge "Wrap macro definition in do-while(0)"
Jerome Jiang [Wed, 20 Mar 2019 17:27:36 +0000 (17:27 +0000)]
Merge "Wrap macro definition in do-while(0)"

5 years agoMerge "Add block size scaling in rd_variance_adjustment()"
Paul Wilkins [Wed, 20 Mar 2019 11:41:39 +0000 (11:41 +0000)]
Merge "Add block size scaling in rd_variance_adjustment()"

5 years agoMerge "Changes to rd_variance_adjustment()"
Paul Wilkins [Wed, 20 Mar 2019 11:41:17 +0000 (11:41 +0000)]
Merge "Changes to rd_variance_adjustment()"

5 years agoMerge "Change to mode early breakout rules for FILM mode."
Paul Wilkins [Wed, 20 Mar 2019 11:41:01 +0000 (11:41 +0000)]
Merge "Change to mode early breakout rules for FILM mode."

5 years agoMerge "Store a group level noise metric from first pass."
Paul Wilkins [Wed, 20 Mar 2019 11:40:46 +0000 (11:40 +0000)]
Merge "Store a group level noise metric from first pass."

5 years agoMerge "vp9-rtc: Adjustments for base layer in screen."
Marco Paniconi [Wed, 20 Mar 2019 03:23:51 +0000 (03:23 +0000)]
Merge "vp9-rtc: Adjustments for base layer in screen."

5 years agoWrap macro definition in do-while(0)
Jerome Jiang [Wed, 20 Mar 2019 00:16:41 +0000 (17:16 -0700)]
Wrap macro definition in do-while(0)

Change-Id: Id654a48d2fa40355552d7267e58461e6cc1c6998

5 years agovp9-rtc: Adjustments for base layer in screen.
Marco Paniconi [Fri, 15 Mar 2019 16:13:45 +0000 (09:13 -0700)]
vp9-rtc: Adjustments for base layer in screen.

On scene/content changes for base layer of screen:
reduce 32x32 split threshold, bias rdcost for flat
blocks if sse_y is non-zero, and avoid early exit on
intra-check.

Reduce artifacts in scroll content.

Change-Id: I144357a61462351173af900e0b8a47dac4aad6ca

5 years agovp9: remove condition on high bitdepth using simple block yrd.
Jerome Jiang [Tue, 19 Mar 2019 23:06:23 +0000 (16:06 -0700)]
vp9: remove condition on high bitdepth using simple block yrd.

Speed is similar between non HBD vs 8bit with HBD build.

BUG=webm:1541

Change-Id: I8b5f7eff87ec7dc4710d31744155a60e50b0f0a9

5 years agoMerge "Compute count_ls in vp9_kmeans"
Angie Chiang [Tue, 19 Mar 2019 18:41:46 +0000 (18:41 +0000)]
Merge "Compute count_ls in vp9_kmeans"

5 years agoNormalize the Wiener variance for ranking
Jingning Han [Mon, 18 Mar 2019 20:52:05 +0000 (13:52 -0700)]
Normalize the Wiener variance for ranking

Normalize the Wiener variance calculation for stack ranking. Remove
potential dependency on blocks at frame boundary.

Change-Id: I37e8634d714a1c34e99f9f7c4f1bb6ea81d56112

5 years agoAdd perceptual AQ mode control to RD search
Jingning Han [Mon, 18 Mar 2019 17:01:11 +0000 (10:01 -0700)]
Add perceptual AQ mode control to RD search

Make the rate-distortion optimization search support the perceptual
quality AQ mode.

Change-Id: Iee507ccfda90ac39b3623de705f187b1459e57e1

5 years agoMerge "Add rdmult adjustment for perceptual AQ mode"
Jingning Han [Mon, 18 Mar 2019 22:30:17 +0000 (22:30 +0000)]
Merge "Add rdmult adjustment for perceptual AQ mode"

5 years agoCompute count_ls in vp9_kmeans
Angie Chiang [Sat, 16 Mar 2019 01:30:38 +0000 (18:30 -0700)]
Compute count_ls in vp9_kmeans

Change-Id: Id5a83554f2037b03a3a7d86f83e47cac311fbe1d

5 years agoMerge "Fix race condition in wiener_var_rd_mult"
Angie Chiang [Mon, 18 Mar 2019 19:39:04 +0000 (19:39 +0000)]
Merge "Fix race condition in wiener_var_rd_mult"

5 years agoAdd rdmult adjustment for perceptual AQ mode
Jingning Han [Sat, 16 Mar 2019 00:02:43 +0000 (17:02 -0700)]
Add rdmult adjustment for perceptual AQ mode

Compute the Lagrangian multiplier for the adaptive quantization
settings.

Change-Id: Ieebe074d6f8163e7541264cb0ead22432273e338

5 years agoMerge "Improve key frame detection"
Jingning Han [Mon, 18 Mar 2019 14:11:18 +0000 (14:11 +0000)]
Merge "Improve key frame detection"

5 years agoImprove key frame detection
Paul Wilkins [Wed, 13 Mar 2019 16:20:23 +0000 (16:20 +0000)]
Improve key frame detection

Improve detection of key frames especially in low contrast
and low motion regions.

This patch adds a function to the key frame detection to test
for specific patterns in the intra signal in the first pass stats
that tend to be indicative of a key frame.

This is intended to compliment the existing code and finds some
scene cuts that were previously being misssed.

Tested on two clips where the existing code was struggling to
identify the key frames this patch improved detection as follows.

Film clip 1: (detected / actual)
Old (2/5) New (5/5)

Film Clip 2
Old 4/11 and one false +,  New 7/11 and 1 false +.

Short 4K Film Scene
Old 1/2 New 2/2

In testing so far I have not seen many extra false +'s though
it is likely that there will be some cases and this may need
further tweaking.

One one of our longer form film test reels ~20k frames)
the change picked up around 35 key frames that were
previously missed, mainly in darker scenes. There were a few
extra (or different) false positives cause by bright flashes or
explosions but these were cases where there was little
difference between inter and intra coding.

Awaiting testing on standard sets.

Change-Id: I1ff4a587e0a47667eb93b197f39b79a1130faeca

5 years agoMerge "Unify the rd_mult use in rd_pick_partition"
Jingning Han [Mon, 18 Mar 2019 04:03:38 +0000 (04:03 +0000)]
Merge "Unify the rd_mult use in rd_pick_partition"

5 years agoMerge "Setup AQ mode for perceptual quality"
Jingning Han [Mon, 18 Mar 2019 04:03:28 +0000 (04:03 +0000)]
Merge "Setup AQ mode for perceptual quality"

5 years agoFix race condition in wiener_var_rd_mult
Angie Chiang [Thu, 14 Mar 2019 19:28:25 +0000 (12:28 -0700)]
Fix race condition in wiener_var_rd_mult

Change-Id: Id5e9c2cbfe35809ac99a3bc9ba93cf462a6b1a34

5 years agoUnify the rd_mult use in rd_pick_partition
Jingning Han [Fri, 15 Mar 2019 23:23:43 +0000 (16:23 -0700)]
Unify the rd_mult use in rd_pick_partition

Abstract the control outside rd_pick_partition function. No need
to switch between x->cb_rdmult and the cpi->rd.RDMULT here.

Change-Id: Ia3104ebe15b5e59a4f29ffe6e8c7d718ecb998a8

5 years agoMerge "vp9-rtc: Some adjustments for low-resolns real-time"
Marco Paniconi [Fri, 15 Mar 2019 23:10:50 +0000 (23:10 +0000)]
Merge "vp9-rtc: Some adjustments for low-resolns real-time"

5 years agoSetup AQ mode for perceptual quality
Jingning Han [Fri, 15 Mar 2019 18:42:39 +0000 (11:42 -0700)]
Setup AQ mode for perceptual quality

Adapt the quantization to provide higher quality at smooth regions
where the Wiener variance is smaller.

Change-Id: Ibfd594d1de2ba34d2440d0aa7991b0fdac057ea5

5 years agoMerge "Refactor tile boundary condition for intra prediction"
Jingning Han [Fri, 15 Mar 2019 22:43:20 +0000 (22:43 +0000)]
Merge "Refactor tile boundary condition for intra prediction"

5 years agovp9-rtc: Some adjustments for low-resolns real-time
Marco Paniconi [Fri, 15 Mar 2019 00:27:58 +0000 (17:27 -0700)]
vp9-rtc: Some adjustments for low-resolns real-time

Force smooth_interpol filter for low resolutions at high Q,
avoid the loopfilter strength reduction for similar conditon,
and reduce thresh_motion for cyclic refresh turnoff.

Change-Id: I4e9121d1cdc7d1b04992c741dc4f0cec281592f7

5 years agoMerge "Setup frame segmentation for perceptual quality target"
Jingning Han [Fri, 15 Mar 2019 16:47:43 +0000 (16:47 +0000)]
Merge "Setup frame segmentation for perceptual quality target"

5 years agoset_analyzer_env.sh: improve cfi diagnostics
James Zern [Fri, 15 Mar 2019 06:07:23 +0000 (23:07 -0700)]
set_analyzer_env.sh: improve cfi diagnostics

use -fno-sanitize-trap=cfi to allow a diagnostic to be printed rather
than aborting with a SIGILL.

https://clang.llvm.org/docs/ControlFlowIntegrity.html#trapping-and-diagnostics

Change-Id: I4517cafe3c7b7305ba4845dbadf9fb679c686843

5 years agoSetup frame segmentation for perceptual quality target
Jingning Han [Fri, 15 Mar 2019 03:26:49 +0000 (20:26 -0700)]
Setup frame segmentation for perceptual quality target

Build the frame segmentation for perceptual quality target.

Change-Id: Icdea28aea02d84ce9c8011f63a3f97133f94a141

5 years agoMerge "Let vp9_kmeans provide boundary for each group"
Jingning Han [Fri, 15 Mar 2019 03:24:23 +0000 (03:24 +0000)]
Merge "Let vp9_kmeans provide boundary for each group"

5 years agoLet vp9_kmeans provide boundary for each group
Angie Chiang [Thu, 14 Mar 2019 22:06:45 +0000 (15:06 -0700)]
Let vp9_kmeans provide boundary for each group

boundary_ls[j] is the upper bound of data centered at ctr_ls[j]

Add vp9_get_group_idx() for computing group_idx

Change-Id: I3b1b488edf8acbfb63c469eeeba15f3e42b0a645

5 years agoMerge "Enclose macro arguments in parentheses"
Jerome Jiang [Thu, 14 Mar 2019 23:08:43 +0000 (23:08 +0000)]
Merge "Enclose macro arguments in parentheses"

5 years agoRefactor tile boundary condition for intra prediction
Jingning Han [Thu, 14 Mar 2019 15:28:44 +0000 (08:28 -0700)]
Refactor tile boundary condition for intra prediction

Explicitly compare the block location against tile coordinate to
decide if intra prediction boundary is available. No coding stats
will be changed by this refactoring.

Change-Id: I80b3a131366bb2c5f8ea53a139ed6e9b0b7ddb68

5 years agoEnclose macro arguments in parentheses
Jerome Jiang [Wed, 13 Mar 2019 23:21:43 +0000 (16:21 -0700)]
Enclose macro arguments in parentheses

BUG=webm:1606

Change-Id: I661485b860243c95b6450035dbac77b0dd4d9ff4

5 years agoMerge changes Ifc006890,I753920a6
Angie Chiang [Thu, 14 Mar 2019 17:22:09 +0000 (17:22 +0000)]
Merge changes Ifc006890,I753920a6

* changes:
  Apply kmeans on log of wiener_variance
  Add vp9_kmeans()

5 years agoMerge "vp9-svc: Reorganize the simulcast mode"
Marco Paniconi [Thu, 14 Mar 2019 04:51:46 +0000 (04:51 +0000)]
Merge "vp9-svc: Reorganize the simulcast mode"

5 years agoMerge "vp9-rtc: Avoid TM intra on big blocks for screen"
Marco Paniconi [Thu, 14 Mar 2019 01:59:49 +0000 (01:59 +0000)]
Merge "vp9-rtc: Avoid TM intra on big blocks for screen"

5 years agoApply kmeans on log of wiener_variance
Angie Chiang [Tue, 12 Mar 2019 22:50:04 +0000 (15:50 -0700)]
Apply kmeans on log of wiener_variance

Change-Id: Ifc0068902e036cd5d94fc5d39f0349a5875469b7

5 years agoAdd vp9_kmeans()
Angie Chiang [Tue, 12 Mar 2019 01:40:53 +0000 (18:40 -0700)]
Add vp9_kmeans()

Change-Id: I753920a65fb14a252617444f49feb40dc332d766

5 years agoMerge "Safe guard zero median filter outcome case"
Angie Chiang [Thu, 14 Mar 2019 00:18:04 +0000 (00:18 +0000)]
Merge "Safe guard zero median filter outcome case"

5 years agoMerge "Refactor speed feature settings"
Angie Chiang [Thu, 14 Mar 2019 00:13:47 +0000 (00:13 +0000)]
Merge "Refactor speed feature settings"

5 years agovp9-rtc: Avoid TM intra on big blocks for screen
Marco Paniconi [Wed, 13 Mar 2019 22:55:18 +0000 (15:55 -0700)]
vp9-rtc: Avoid TM intra on big blocks for screen

For screen content real-time mode: don't check TM
intra for bsize >= BLOCK_32X32.

Small speedup and avoid some artifacts seen
in scrolling screen content.

Change-Id: I72d7731eeb6ac9ee96e65af522c1a9aabb6dc4ef

5 years agovp9-svc: Reorganize the simulcast mode
Marco Paniconi [Wed, 13 Mar 2019 20:46:12 +0000 (13:46 -0700)]
vp9-svc: Reorganize the simulcast mode

Set the lst/gld/alt_fb_idx and refresh flags for
key frames at the start of encoding (in svc_set_params).
This then avoids new code/function in update_references()
and in copy_flags_ref_update().

Change-Id: Id3503c0c628540c20f11a540c118c4ee4cf04848

5 years agoMerge "vp9 simulcast: update buffer slot flag used in API."
Marco Paniconi [Wed, 13 Mar 2019 20:16:17 +0000 (20:16 +0000)]
Merge "vp9 simulcast: update buffer slot flag used in API."

5 years agoSafe guard zero median filter outcome case
Jingning Han [Wed, 13 Mar 2019 19:06:39 +0000 (12:06 -0700)]
Safe guard zero median filter outcome case

In case median filter returns 0 value, bypass the Wiener filter
stage. This avoids potential divided by 0 case.

In the same place use a temp variable to take the Wiener filter
output instead of returning to the coeff array.

Change-Id: I45f57c515b4062a0aa1f312eda852462cb655d8e

5 years agovp9 simulcast: update buffer slot flag used in API.
Jerome Jiang [Wed, 13 Mar 2019 16:13:35 +0000 (09:13 -0700)]
vp9 simulcast: update buffer slot flag used in API.

Also add #spatial_layers > 1 to simulcast_mode

Change-Id: I6234da81801176ac8678f9f5e1323f8b289cb663

5 years agoRefactor speed feature settings
Jingning Han [Wed, 13 Mar 2019 18:20:28 +0000 (11:20 -0700)]
Refactor speed feature settings

Make the speed feature setup functions take speed argument as
their input.

Change-Id: I542e8f6e04658e5d99e972380a31baab99a4fc23

5 years agoAdaptive multiplier based on the Wiener variance
Jingning Han [Wed, 6 Mar 2019 01:06:20 +0000 (17:06 -0800)]
Adaptive multiplier based on the Wiener variance

Adapt the Lagrangian multiplier based on the Wiener variance at
64x64 block level.

Change-Id: Ica195ed6f706daf6eee156d4b1a55bda65a92f7b

5 years agoMerge "Add normalization over frame level Wiener variance"
Jingning Han [Wed, 13 Mar 2019 16:18:09 +0000 (16:18 +0000)]
Merge "Add normalization over frame level Wiener variance"

5 years agoMerge "Set up Wiener variance for macroblocks in a frame"
Jingning Han [Wed, 13 Mar 2019 16:17:52 +0000 (16:17 +0000)]
Merge "Set up Wiener variance for macroblocks in a frame"

5 years agoAdd normalization over frame level Wiener variance
Jingning Han [Tue, 5 Mar 2019 20:14:40 +0000 (12:14 -0800)]
Add normalization over frame level Wiener variance

Normalize the block level Wiener variance based decision according
to the frame level Wiener variance.

Change-Id: Ic2bdf1b322a65661775541dd6c174ba71579461a

5 years agoSet up Wiener variance for macroblocks in a frame
Jingning Han [Tue, 5 Mar 2019 18:45:16 +0000 (10:45 -0800)]
Set up Wiener variance for macroblocks in a frame

This commit introduces a Wiener variance term. For each block in
the source frame, we first estimate its film grain noise level
using median filter in the transform domain. Each transform
coefficient is then processed using Wiener filter to account for
the impact on the energy level due to film grain noise. The result
leads to a second moment of the denoised signal.

Change-Id: Ibce7cb1b0cb8fe1aba807d95289712271d576948

5 years agoRemove highbd_temporal_filter_sse4.c when REAL_TIME_ONLY is on
chiyotsai [Tue, 12 Mar 2019 17:03:00 +0000 (10:03 -0700)]
Remove highbd_temporal_filter_sse4.c when REAL_TIME_ONLY is on

Change-Id: I3ca7442b10cbea4dd5dbabe147687d1cb3cce4d8

5 years agoMerge "vp9-decoder: use long int for buffer offset."
Jerome Jiang [Tue, 12 Mar 2019 00:55:34 +0000 (00:55 +0000)]
Merge "vp9-decoder: use long int for buffer offset."

5 years agoMerge "vp9: map speed > 9 to speed 9."
Jerome Jiang [Mon, 11 Mar 2019 22:45:47 +0000 (22:45 +0000)]
Merge "vp9: map speed > 9 to speed 9."

5 years agovp9-decoder: use long int for buffer offset.
Jerome Jiang [Mon, 11 Mar 2019 22:13:19 +0000 (15:13 -0700)]
vp9-decoder: use long int for buffer offset.

integer overflow when frame size too big.

BUG=webm:1603

Change-Id: Ifbb81b5fb6a2043d09d403e7c50ab8d7bf125dca

5 years agoAdd block size scaling in rd_variance_adjustment()
Paul Wilkins [Thu, 28 Feb 2019 16:58:19 +0000 (16:58 +0000)]
Add block size scaling in rd_variance_adjustment()

Scale the block variance values used in this function
to a common block size.

Change-Id: I73ad7d48b2621f312d771ee0dd7b6fc59cfc1652

5 years agoChanges to rd_variance_adjustment()
Paul Wilkins [Thu, 28 Feb 2019 16:23:35 +0000 (16:23 +0000)]
Changes to rd_variance_adjustment()

Always calculate per block variance values vs per pixel.

Change-Id: I760b3ba1a250d7544813a1b93923eedc207cbd60

5 years agoChange to mode early breakout rules for FILM mode.
Paul Wilkins [Thu, 28 Feb 2019 14:10:12 +0000 (14:10 +0000)]
Change to mode early breakout rules for FILM mode.

The biggest offender in terms of preventing retention of film
grain in high rate film content is the use of DC-PRED mode.
Some of the directional modes whilst not strictly preserving
grain do better at at least preserving some texture.

This change blocks the early breakout of the rd loop based
on the reference frame giving the best result so far. In practice,
unless DC-PRED was chosen as the best mode so far, the other
directional intra modes would not even be considered.

As the film grain mode also tends to bias against DC-PRED (or
intra in general) this was pretty much blocking all use of directional
intra modes.

The patch also allows for a broader spectrum of DC modes at the
16x16 transform level than previously.

Change-Id: I860b7726ea9f5fcbb3ec1a90edbdd8cade2e8b28

5 years agoStore a group level noise metric from first pass.
Paul Wilkins [Thu, 28 Feb 2019 13:19:11 +0000 (13:19 +0000)]
Store a group level noise metric from first pass.

Adds a variable to GF group structure to store a noise
energy metric for the current group that can be used
in things like film grain retention code.

Change-Id: I81b07630d3242f7928110f19a6c1ed4c86125f05

5 years agovp9-screen: Rework the mode skip logic for screen
Marco Paniconi [Sat, 9 Mar 2019 05:00:48 +0000 (21:00 -0800)]
vp9-screen: Rework the mode skip logic for screen

Don't force skip of zero-golden reference when
zero_temp_sad_source = 0, as it be may the
inter-layer reference. And remove the flatness conditon
when superblock is static.

Change-Id: I6b4b6eac0f6a2abc862c23d0e5467c7cf61995ef