External denoiser based on noise estimation and moving object detection.

External denoiser based on noise estimation and moving object detection.

Improved the existing external denoiser in WebRTC: the filter strength
is adaptive based on the noise level of the whole frame and the moving
object detection result. The adaptive filter effectively removes the
artifacts in previous version, such as trailing and blockiness on moving
objects.
The external denoiser is off by default for now.

BUG=

Committed: https://crrev.com/fa0befe13bb030adb61eb55c5557df1bf6e29205
Cr-Commit-Position: refs/heads/master@{#12198}

[jackychen_, 2016-03-24 01:14:50]

Patchset #1 (id:1) has been deleted

[jackychen_, 2016-03-24 01:15:00]

Patchset #1 (id:20001) has been deleted

[jackychen_, 2016-03-24 01:23:39]

Description was changed from

==========
External denoiser based on noise estimation and moving object detection.

Improved the existing external denoiser in WebRTC: the filter strength
is adaptive based on the noise level of the whole frame and the moving
object detection result. The adaptive filter effectively removes the
artifacts in previous version, such as trailing and blockyness on moving
objects.
The external denoiser is off by default for now.

BUG=
==========

to

==========
External denoiser based on noise estimation and moving object detection.

Improved the existing external denoiser in WebRTC: the filter strength
is adaptive based on the noise level of the whole frame and the moving
object detection result. The adaptive filter effectively removes the
artifacts in previous version, such as trailing and blockyness on moving
objects.
The external denoiser is off by default for now.

BUG=
==========

[jackychen_, 2016-03-24 03:47:56]

jackychen@google.com changed reviewers:
+ marpan@webrtc.org

[jackychen_, 2016-03-24 03:48:32]

jackychen@google.com changed reviewers:
+ jackychen@google.com

[marpan, 2016-03-24 20:56:15]

marpan@google.com changed reviewers:
+ marpan@google.com

[marpan, 2016-03-24 20:56:17]

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
File webrtc/modules/video_processing/BUILD.gn (right):

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/BUILD.gn:26: "noise_estimation.cc",
put these under /util

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
File webrtc/modules/video_processing/frame_preprocessor.h (right):

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/frame_preprocessor.h:74: VideoFrame
denoised_frame_track_;
is this previous denoised_frame (i.e, t-1)?
if so, maybe use the name denoised_frame_prev

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
File webrtc/modules/video_processing/noise_estimation.cc (right):

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/noise_estimation.cc:26: if
(consec_low_var_[mb_index] >= 6 && (luma >> 8) < 220 && (luma >> 8) > 20) {
i would prefer to expose these thresholds (6, 229, 290, 384) as parameters
of model, with name that explains its meaning.

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/noise_estimation.cc:39: // No enough samples
implies the motion of the camera.
...or too much of object in frame

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/noise_estimation.cc:65: noise_var_accum_ =
(noise_var_accum_ * 15 + noise_var_) / 16;
better to expose this (15) as parameter too?

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
File webrtc/modules/video_processing/util/denoiser_filter_c.cc (right):

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/util/denoiser_filter_c.cc:142: sum_diff_thresh =
kSumDiffThresholdMax;
maybe don't need to define kSumDiffThresholdMax, just do:
if (denoise_always)
 sum_diff_thresh = INT_MAX

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
File webrtc/modules/video_processing/video_denoiser.cc (right):

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/video_denoiser.cc:97: for (int mb_col = 0;
mb_col < mb_cols; ++mb_col) {
can you use memset here?

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/video_denoiser.cc:100: for (int mb_row = 0;
mb_row < mb_rows; ++mb_row) {
put more comments to explain what this block of code does.
i.e, something like: "...Loop over blocks, compute variance between ?? and ??,
accumulate/extract noise level, and update x/y_density factors for moving object
extraction...."

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/video_denoiser.cc:144: // Track averaged frame.
Track denoised frame?

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/video_denoiser.cc:200: for (int mb_row = 0;
mb_row < mb_rows; ++mb_row) {
put comments to explain this block of code

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
File webrtc/modules/video_processing/video_processing.gypi (right):

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/video_processing.gypi:31: 'noise_estimation.cc',
should these be put in /util/noise_estimation

[jackychen_, 2016-03-25 18:45:45]

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
File webrtc/modules/video_processing/BUILD.gn (right):

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/BUILD.gn:26: "noise_estimation.cc",
On 2016/03/24 20:56:16, marpan wrote:
> put these under /util

Done.

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
File webrtc/modules/video_processing/frame_preprocessor.h (right):

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/frame_preprocessor.h:74: VideoFrame
denoised_frame_track_;
On 2016/03/24 20:56:16, marpan wrote:
> is this previous denoised_frame (i.e, t-1)?
> if so, maybe use the name denoised_frame_prev

Done.

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
File webrtc/modules/video_processing/noise_estimation.cc (right):

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/noise_estimation.cc:26: if
(consec_low_var_[mb_index] >= 6 && (luma >> 8) < 220 && (luma >> 8) > 20) {
On 2016/03/24 20:56:16, marpan wrote:
> i would prefer to expose these thresholds (6, 229, 290, 384) as parameters
> of model, with name that explains its meaning.

Done.

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/noise_estimation.cc:39: // No enough samples
implies the motion of the camera.
On 2016/03/24 20:56:16, marpan wrote:
> ...or too much of object in frame

Done.

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/noise_estimation.cc:65: noise_var_accum_ =
(noise_var_accum_ * 15 + noise_var_) / 16;
On 2016/03/24 20:56:16, marpan wrote:
> better to expose this (15) as parameter too?

Acknowledged.

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
File webrtc/modules/video_processing/util/denoiser_filter_c.cc (right):

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/util/denoiser_filter_c.cc:142: sum_diff_thresh =
kSumDiffThresholdMax;
On 2016/03/24 20:56:16, marpan wrote:
> maybe don't need to define kSumDiffThresholdMax, just do:
> if (denoise_always)
>  sum_diff_thresh = INT_MAX

Done.

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
File webrtc/modules/video_processing/video_denoiser.cc (right):

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/video_denoiser.cc:97: for (int mb_col = 0;
mb_col < mb_cols; ++mb_col) {
On 2016/03/24 20:56:16, marpan wrote:
> can you use memset here?

Done.

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/video_denoiser.cc:100: for (int mb_row = 0;
mb_row < mb_rows; ++mb_row) {
On 2016/03/24 20:56:16, marpan wrote:
> put more comments to explain what this block of code does.
> i.e, something like: "...Loop over blocks, compute variance between ?? and ??,
> accumulate/extract noise level, and update x/y_density factors for moving
object
> extraction...."

Done.

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/video_denoiser.cc:144: // Track averaged frame.
On 2016/03/24 20:56:16, marpan wrote:
> Track denoised frame?

Done.

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/video_denoiser.cc:200: for (int mb_row = 0;
mb_row < mb_rows; ++mb_row) {
On 2016/03/24 20:56:16, marpan wrote:
> put comments to explain this block of code

Done.

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
File webrtc/modules/video_processing/video_processing.gypi (right):

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/video_processing.gypi:31: 'noise_estimation.cc',
On 2016/03/24 20:56:17, marpan wrote:
> should these be put in /util/noise_estimation

Done.

[marpan, 2016-03-30 16:14:26]

Description was changed from

==========
External denoiser based on noise estimation and moving object detection.

Improved the existing external denoiser in WebRTC: the filter strength
is adaptive based on the noise level of the whole frame and the moving
object detection result. The adaptive filter effectively removes the
artifacts in previous version, such as trailing and blockyness on moving
objects.
The external denoiser is off by default for now.

BUG=
==========

to

==========
External denoiser based on noise estimation and moving object detection.

Improved the existing external denoiser in WebRTC: the filter strength
is adaptive based on the noise level of the whole frame and the moving
object detection result. The adaptive filter effectively removes the
artifacts in previous version, such as trailing and blockiness on moving
objects.
The external denoiser is off by default for now.

BUG=
==========

[marpan, 2016-03-30 18:27:20]

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
File webrtc/modules/video_processing/video_denoiser.cc (right):

https://codereview.webrtc.org/1822333003/diff/40001/webrtc/modules/video_proc...
webrtc/modules/video_processing/video_denoiser.cc:119: // Get the averaged
block.
Get denoised block

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
File webrtc/modules/video_processing/util/denoiser_filter.h (right):

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/denoiser_filter.h:25: extern const int
kSumDiffThresholdMax;
do we need this parameter?

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
File webrtc/modules/video_processing/util/denoiser_filter_c.cc (right):

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/denoiser_filter_c.cc:141: if
(denoise_always)
better to put this before 140?
if (denoise_always) {
  sum_diff_thresh = INT_MAX;
}else {
  if (increase_denoising)
     sum_diff_thresh = kSumDiffThresholdHigh;
  else
    sum_diff_thresh = kSumDiffThreshold;
}

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
File webrtc/modules/video_processing/util/denoiser_filter_neon.cc (right):

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/denoiser_filter_neon.cc:198: if
(denoise_always)
same comment as in _c code

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
File webrtc/modules/video_processing/util/denoiser_filter_sse2.cc (right):

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/denoiser_filter_sse2.cc:216: if
(denoise_always)
same comment as _c code

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
File webrtc/modules/video_processing/util/noise_estimation.cc (right):

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:24: void
NoiseEstimation::GetNoise(int mb_index, uint32_t var, uint32_t luma) {
Comments: This computes block noise and accumulates?

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:40: uint8_t
NoiseEstimation::GetNoiseLevel() {
You're still doing 3 things in this function:
1. normalize the frame-level noise estimate
2. update with recursive average
3. and threshold to get noise level

Still feel splitting this into at least 2 functions would be better.

Also some comments to what function(s) does.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:42: int mb_rows =
height_ >> 4;
is it worth defining mb_cols/rows_ in this class, so only need to set once in
Init.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:43: // No enough
samples implies the motion of the camera or too many moving
Should condition of "Not enough samples" be num_noisy_block > thresh ?

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:45: if
(num_stationary_block_ < 0.65 * mb_cols * mb_rows || !num_noisy_block_) {
prefer  (0.65 * mb_cols * mb_rows)

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:47: noise_var_accum_ =
0;
is this intentional to set (the running average) noise_var_accum_ = 0 if a frame
suddenly has alot of motion?

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:49: num_noisy_block_ =
0;
the 3 quantities (noise_var, num_stationary_block_, num_noisy_block_) seem to
get reset
anyway below. Maybe can we do it only once at end of function?

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:56:
percent_non_montion_block_ =
non_motion

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
File webrtc/modules/video_processing/video_denoiser.cc (right):

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/video_denoiser.cc:155: // Draft. This can be
optimized. This code block is to reduce false detection
is it better to put this code block into separate function, i.e, 
ReduceFalseDetection()

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/video_denoiser.cc:240: #if DISPLAY  // Rectangle
diagnostics
good to put this "display code" into separate function.

[jackychen_, 2016-03-30 22:33:21]

The CQ bit was checked by jackychen@google.com

[jackychen_, 2016-03-30 22:33:22]

The CQ bit was unchecked by jackychen@google.com

[jackychen_, 2016-03-30 22:36:38]

Fixed.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
File webrtc/modules/video_processing/util/denoiser_filter.h (right):

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/denoiser_filter.h:25: extern const int
kSumDiffThresholdMax;
On 2016/03/30 18:27:20, marpan wrote:
> do we need this parameter?

Done.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
File webrtc/modules/video_processing/util/denoiser_filter_c.cc (right):

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/denoiser_filter_c.cc:141: if
(denoise_always)
On 2016/03/30 18:27:20, marpan wrote:
> better to put this before 140?
> if (denoise_always) {
>   sum_diff_thresh = INT_MAX;
> }else {
>   if (increase_denoising)
>      sum_diff_thresh = kSumDiffThresholdHigh;
>   else
>     sum_diff_thresh = kSumDiffThreshold;
> }

Done.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
File webrtc/modules/video_processing/util/denoiser_filter_neon.cc (right):

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/denoiser_filter_neon.cc:198: if
(denoise_always)
On 2016/03/30 18:27:20, marpan wrote:
> same comment as in _c code

Done.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
File webrtc/modules/video_processing/util/denoiser_filter_sse2.cc (right):

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/denoiser_filter_sse2.cc:216: if
(denoise_always)
On 2016/03/30 18:27:20, marpan wrote:
> same comment as _c code

Done.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
File webrtc/modules/video_processing/util/noise_estimation.cc (right):

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:24: void
NoiseEstimation::GetNoise(int mb_index, uint32_t var, uint32_t luma) {
On 2016/03/30 18:27:20, marpan wrote:
> Comments: This computes block noise and accumulates?

It normalizes var and accumulates. var is computed outside, I actually want to
move it in to save some cycles (if we do noise estimation less frequent, we save
one variance computation). Since it is related to speed-up, I postpone it to
next cl.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:40: uint8_t
NoiseEstimation::GetNoiseLevel() {
On 2016/03/30 18:27:20, marpan wrote:
> You're still doing 3 things in this function:
> 1. normalize the frame-level noise estimate
> 2. update with recursive average
> 3. and threshold to get noise level
> 
> Still feel splitting this into at least 2 functions would be better.
> 
> Also some comments to what function(s) does.

Split to 2 functions and add some comments.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:42: int mb_rows =
height_ >> 4;
On 2016/03/30 18:27:20, marpan wrote:
> is it worth defining mb_cols/rows_ in this class, so only need to set once in
> Init.

Done.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:43: // No enough
samples implies the motion of the camera or too many moving
On 2016/03/30 18:27:20, marpan wrote:
> Should condition of "Not enough samples" be num_noisy_block > thresh ?

Correct. I just use thresh = 0, so I code as !num_noisy_block. It's actually a
weak condition, but works well with the first part of the condition.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:45: if
(num_stationary_block_ < 0.65 * mb_cols * mb_rows || !num_noisy_block_) {
On 2016/03/30 18:27:20, marpan wrote:
> prefer  (0.65 * mb_cols * mb_rows)

Done.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:47: noise_var_accum_ =
0;
On 2016/03/30 18:27:20, marpan wrote:
> is this intentional to set (the running average) noise_var_accum_ = 0 if a
frame
> suddenly has alot of motion?

Yes. When it happens, I think we should start over. The logic may change if we
compute noise level less frequent, say every 10 frames.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:49: num_noisy_block_ =
0;
On 2016/03/30 18:27:20, marpan wrote:
> the 3 quantities (noise_var, num_stationary_block_, num_noisy_block_) seem to
> get reset
> anyway below. Maybe can we do it only once at end of function?

There is a "return" in the last line of if branch, not easy to find. So need to
keep both.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:56:
percent_non_montion_block_ =
On 2016/03/30 18:27:20, marpan wrote:
> non_motion

Maybe it's better to use "static" to make it concise.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
File webrtc/modules/video_processing/video_denoiser.cc (right):

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/video_denoiser.cc:155: // Draft. This can be
optimized. This code block is to reduce false detection
On 2016/03/30 18:27:20, marpan wrote:
> is it better to put this code block into separate function, i.e, 
> ReduceFalseDetection()

Done.

https://codereview.webrtc.org/1822333003/diff/100001/webrtc/modules/video_pro...
webrtc/modules/video_processing/video_denoiser.cc:240: #if DISPLAY  // Rectangle
diagnostics
On 2016/03/30 18:27:20, marpan wrote:
> good to put this "display code" into separate function.

Done.

[marpan, 2016-03-31 18:36:47]

lgtm

https://codereview.webrtc.org/1822333003/diff/120001/webrtc/modules/video_pro...
File webrtc/modules/video_processing/util/noise_estimation.cc (right):

https://codereview.webrtc.org/1822333003/diff/120001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:31: // Normalized var
by the average luma value, this gives more credit to
weight?

https://codereview.webrtc.org/1822333003/diff/120001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:47: if
(num_static_block_ < (0.65 * mb_cols_ * mb_rows_) || !num_noisy_block_) {
look into why not "num_noisy_block_ < T"

[jackychen_, 2016-03-31 22:33:31]

Patchset #6 (id:140001) has been deleted

[jackychen_, 2016-04-01 00:05:06]

https://codereview.webrtc.org/1822333003/diff/120001/webrtc/modules/video_pro...
File webrtc/modules/video_processing/util/noise_estimation.cc (right):

https://codereview.webrtc.org/1822333003/diff/120001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:31: // Normalized var
by the average luma value, this gives more credit to
On 2016/03/31 18:36:47, marpan wrote:
> weight?

Done.

https://codereview.webrtc.org/1822333003/diff/120001/webrtc/modules/video_pro...
webrtc/modules/video_processing/util/noise_estimation.cc:47: if
(num_static_block_ < (0.65 * mb_cols_ * mb_rows_) || !num_noisy_block_) {
On 2016/03/31 18:36:47, marpan wrote:
> look into why not "num_noisy_block_ < T"

Done.

[marpan2, 2016-04-01 00:06:02]

lgtm

[jackychen_, 2016-04-01 14:45:11]

The CQ bit was checked by jackychen@google.com

[jackychen_, 2016-04-01 14:45:11]

The patchset sent to the CQ was uploaded after l-g-t-m from marpan@google.com
Link to the patchset: https://codereview.webrtc.org/1822333003/#ps160001 (title:
" ")

[commit-bot: I haz the power, 2016-04-01 14:45:23]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1822333003/160001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1822333003/160001

[commit-bot: I haz the power, 2016-04-01 14:47:03]

Description was changed from

==========
External denoiser based on noise estimation and moving object detection.

Improved the existing external denoiser in WebRTC: the filter strength
is adaptive based on the noise level of the whole frame and the moving
object detection result. The adaptive filter effectively removes the
artifacts in previous version, such as trailing and blockiness on moving
objects.
The external denoiser is off by default for now.

BUG=
==========

to

==========
External denoiser based on noise estimation and moving object detection.

Improved the existing external denoiser in WebRTC: the filter strength
is adaptive based on the noise level of the whole frame and the moving
object detection result. The adaptive filter effectively removes the
artifacts in previous version, such as trailing and blockiness on moving
objects.
The external denoiser is off by default for now.

BUG=
==========

[commit-bot: I haz the power, 2016-04-01 14:47:03]

Committed patchset #6 (id:160001)

[commit-bot: I haz the power, 2016-04-01 14:47:10]

Description was changed from

==========
External denoiser based on noise estimation and moving object detection.

Improved the existing external denoiser in WebRTC: the filter strength
is adaptive based on the noise level of the whole frame and the moving
object detection result. The adaptive filter effectively removes the
artifacts in previous version, such as trailing and blockiness on moving
objects.
The external denoiser is off by default for now.

BUG=
==========

to

==========
External denoiser based on noise estimation and moving object detection.

Improved the existing external denoiser in WebRTC: the filter strength
is adaptive based on the noise level of the whole frame and the moving
object detection result. The adaptive filter effectively removes the
artifacts in previous version, such as trailing and blockiness on moving
objects.
The external denoiser is off by default for now.

BUG=

Committed: https://crrev.com/fa0befe13bb030adb61eb55c5557df1bf6e29205
Cr-Commit-Position: refs/heads/master@{#12198}
==========

[commit-bot: I haz the power, 2016-04-01 14:47:11]

Patchset 6 (id:??) landed as
https://crrev.com/fa0befe13bb030adb61eb55c5557df1bf6e29205
Cr-Commit-Position: refs/heads/master@{#12198}