★ Record the image dominant color in the image database.

★ Record the image dominant color in the image database.

This CL adds a column in the bookmarks images database.
It contains the dominant color extracted from the image.
Since this computation can be lengthy, it is done in the background, when the
image is stored in the database.
For old databases, a migration path is provided and the dominant color is
computed and stored upon querying the image for the first time.

BUG=448443
Committed: https://crrev.com/a110f52e83cebd2b6f0ed37c34abeac8bfcb1b72
Cr-Commit-Position: refs/heads/master@{#312916}

[lpromero, 2015-01-23 17:48:53]

lpromero@chromium.org changed reviewers:
+ bsalomon@google.com, noyau@chromium.org

[noyau (Ping after 24h), 2015-01-23 17:53:32]

lgtm

CC +kkimlabs

[lpromero, 2015-01-23 17:55:30]

On 2015/01/23 17:54:33, lpromero wrote:

bsalomon@: For context, I added you for the DEPS changes. Skia is now also used
in the component, not just its tests.

[lpromero, 2015-01-23 17:58:21]

Patchset #1 (id:1) has been deleted

[lpromero, 2015-01-23 18:01:37]

On 2015/01/23 17:55:30, lpromero wrote:
> On 2015/01/23 17:54:33, lpromero wrote:
> 
> bsalomon@: For context, I added you for the DEPS changes. Skia is now also
used
> in the component, not just its tests.

+ For context for everyone: this CL was prepared and reviewed in
https://chromereviews.googleplex.com/145967013/ (internal only), hence the quick
LGTM from noyau@ :)

[Kibeom Kim (inactive), 2015-01-23 18:15:37]

kkimlabs@chromium.org changed reviewers:
+ kkimlabs@chromium.org

[bsalomon, 2015-01-23 18:21:48]

On 2015/01/23 18:15:37, Kibeom Kim wrote:

lgtm

[Ian Wen, 2015-01-23 18:22:49]

ianwen@chromium.org changed reviewers:
+ ianwen@chromium.org

[Ian Wen, 2015-01-23 18:22:49]

Question: is computing dominant color a length job to do? On clank we compute
the dominant color on the fly, using getDominantColor() in
chrome/browser/ui/webui/ntp/favicon_webui_handler.cc.
https://chrome-internal-review.googlesource.com/#/c/189561/ . And it seems the
speed is pretty fast on Android.

[lpromero, 2015-01-23 18:40:47]

On 2015/01/23 18:22:49, Ian Wen wrote:
> Question: is computing dominant color a length job to do? On clank we compute
> the dominant color on the fly, using getDominantColor() in
> chrome/browser/ui/webui/ntp/favicon_webui_handler.cc.
> https://chrome-internal-review.googlesource.com/#/c/189561/ . And it seems the
> speed is pretty fast on Android.

Yes, it's the same method underneath: color_utils::CalculateKMeanColorOfBitmap

It took something like 3 seconds on iPhone 6 Plus Release for
http://www.djc.com/blogs/SeattleScape/wp-content/uploads/2012/08/chinatwo.jpg

Otherwise, for small images, it's not noticeable.

[Ian Wen, 2015-01-23 18:57:04]

On 2015/01/23 18:40:47, lpromero wrote:
> On 2015/01/23 18:22:49, Ian Wen wrote:
> > Question: is computing dominant color a length job to do? On clank we
compute
> > the dominant color on the fly, using getDominantColor() in
> > chrome/browser/ui/webui/ntp/favicon_webui_handler.cc.
> > https://chrome-internal-review.googlesource.com/#/c/189561/ . And it seems
the
> > speed is pretty fast on Android.
> 
> Yes, it's the same method underneath: color_utils::CalculateKMeanColorOfBitmap
> 
> It took something like 3 seconds on iPhone 6 Plus Release for
> http://www.djc.com/blogs/SeattleScape/wp-content/uploads/2012/08/chinatwo.jpg
> 
> Otherwise, for small images, it's not noticeable.

Wow that's a large image. Then why don't we crop the image before loading it to
UI? I suppose if we can make images smaller, it would be saving both time and
memory. Correct me if I am wrong. :P

[lpromero, 2015-01-23 19:08:01]

The CQ bit was checked by lpromero@chromium.org

[commit-bot: I haz the power, 2015-01-23 19:10:30]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/875463003/40001

[lpromero, 2015-01-23 19:11:34]

On 2015/01/23 18:57:04, Ian Wen wrote:
> On 2015/01/23 18:40:47, lpromero wrote:
> > On 2015/01/23 18:22:49, Ian Wen wrote:
> > > Question: is computing dominant color a length job to do? On clank we
> compute
> > > the dominant color on the fly, using getDominantColor() in
> > > chrome/browser/ui/webui/ntp/favicon_webui_handler.cc.
> > > https://chrome-internal-review.googlesource.com/#/c/189561/ . And it seems
> the
> > > speed is pretty fast on Android.
> > 
> > Yes, it's the same method underneath:
color_utils::CalculateKMeanColorOfBitmap
> > 
> > It took something like 3 seconds on iPhone 6 Plus Release for
> >
http://www.djc.com/blogs/SeattleScape/wp-content/uploads/2012/08/chinatwo.jpg
> > 
> > Otherwise, for small images, it's not noticeable.
> 
> Wow that's a large image. Then why don't we crop the image before loading it
to
> UI? I suppose if we can make images smaller, it would be saving both time and
> memory. Correct me if I am wrong. :P

You mean resize? Why not. But to which screen size? This image is almost the
same as the
iPhone 6 Plus, so if we crop/resize the image, it's gonna look ugly on the
bookmark editor.

[commit-bot: I haz the power, 2015-01-23 20:18:18]

Committed patchset #2 (id:40001)

[commit-bot: I haz the power, 2015-01-23 20:19:12]

Patchset 2 (id:??) landed as
https://crrev.com/a110f52e83cebd2b6f0ed37c34abeac8bfcb1b72
Cr-Commit-Position: refs/heads/master@{#312916}