Re: [RFC/PATCH 1/5] v4l: Convert V4L2_CID_FOCUS_AUTO control to a menu control

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi Laurent,

On 01/03/2012 02:55 PM, Laurent Pinchart wrote:
>>>>> parameter. We also need to discuss how the af statistics window
>>>>> configuration is done. I'm not certain there could even be a
>>>>> standardised
>>>>
>>>> Do we need multiple windows for AF statistics ?
>>>>
>>>> If not, I'm inclined to use four separate controls for window
>>>> configuration. (X, Y, WIDTH, HEIGHT). This was Hans' preference in
>>>> previous discussions [1].
>>>
>>> For the OMAP3 ISP we need multiple statistics windows. AEWB can use more
>>> than 32 windows. Having separate controls for that wouldn't be
>>> practical.
>>
>> OK, so the control API in current form doesn't seem capable of setting up
>> the statistics windows. There is also little space in struct
>> v4l2_ext_control for any major extensions.
>>
>> We might need to define dedicated set of selection targets in the selection
>> API for handling multiple windows.
>>
>> Yet, to avoid forcing applications to use the selection API where
>> rectangles aren't needed - only single spot coordinates, how about
>> defining following two controls ?
>>
>> * AF spot coordinates when focus mode is set to V4L2_AUTO_FOCUS_MODE_SPOT
>>
>>  - V4L2_CID_AUTO_FOCUS_POSITION_X - horizontal position in pixels relative
>>                                     to the left of frame
>>  - V4L2_CID_AUTO_FOCUS_POSITION_Y - vertical position in pixels relative
>>                                     to the top of frame
> 
> What about a point control type instead ? :-) X and Y coordinates could be 
> stored on 32 bits each.

That's more appealing than two separate controls :-) If Hans agrees to
add a point control type (fingers crossed :)) I could prepare relevant patch
to see how it looks like. I've analysed roughly what would need to be changed,
the effort is quite significant but not so invasive for drivers.

I thought about using new V4L2_CTRL_FLAG* for VIDIOC_QUERYCTRL to indicate
which field of the point data structure is queried.

The only real problem seem to be events, I can't see simple method for adding
two sets of min/max/step/def values to the control event payload. There would
probably have to be two separate control change events for each point structure
field.

-- 

Regards,
Sylwester
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html


[Index of Archives]     [Linux Input]     [Video for Linux]     [Gstreamer Embedded]     [Mplayer Users]     [Linux USB Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [Yosemite Backpacking]
  Powered by Linux