Re: [RFC] Zoom controls in V4L2

[Laurent Pinchart <laurent.pinchart@xxxxxxxxx>]

Hi everybody,

I haven't received any answer so far. Is everybody busy or just not interested 
in zoom support ? Please don't let the length of the mail scare you :-)

On Tuesday 04 November 2008, Laurent Pinchart wrote:
> Hi everybody,
>
> USB cameras with integrated optical zoom support are hitting the market.
> V4L2 currently lacks the necessary controls to support zoom. This RFC tries
> to define zoom-related controls.
>
> As few camera models currently support optical zoom, only a subset of zoom
> functions are implemented in existing products, making it a bit harder to
> define a future proof zoom API in V4L2. To gather more usecases I've taken
> all zoom controls defined in the USB Video Class specification into
> account, even if they are not all implemented in existing products.
>
> Zoom in digital cameras is implemented as optical zoom, digital zoom or a
> combination of both. V4L2 supports digital zoom through cropping and
> scaling (section 1.11). Digital cameras often implement digital zoom
> through a single linear control, providing a subset of the scaling
> capabilities of V4L2 with no easy way to map between both. Still, defining
> a new digital zoom API in addition to the V4L2 cropping and scaling
> mechanism would confuse developers and users and should be avoided. We
> should instead concentrate on defining a clear mapping between linear
> digital zoom and crop/scale.
>
> As I don't own any UVC device with digital zoom support, and as I'm not
> knowledgeable about digital zoom support in non-UVC webcams, ideas for a
> mapping between linear digital zoom and crop/scale are welcome. In case of
> lack of feedback on the subject, I propose to concentrate on optical zoom
> only, except when digital zoom interacts with optical zoom.
>
> The UVC specification approximates the optical magnification factor as the
> ratio between the ocular lens focal length and the objective lens focal
> length. Although lens groups can be much more sophisticated than that, the
> model can approximate most lens groups that are likely to be encountered in
> practice. Zoom can then be expressed either as the magnification factor or
> as the objective lens focal lens. To support both representations V4L2
> should let the device set its minimum and maximum zoom values. In both
> cases the zoom is either an unsigned integer or an unsigned rational number
> that can be expressed with a fixed-point representation.
>
> Optical zoom can be controlled in an absolute or relative fashion. Absolute
> zoom can easily be handled with a single unsigned integer control mapping
> to the magnification factor or objective lens focal length as described
> above. The absolute zoom control should not interact with any digital zoom
> function implemented in the device, if any. I suggest naming the control
> V4L2_CID_ZOOM_ABSOLUTE.
>
> Relative zoom is a tad more complex. To begin with, there are two relative
> zoom implementations I can think of: incremental or continuous. Incremental
> relative zoom moves the optical zoom level by a fixed amount. This is how
> the relative pan, tilt and focus controls are specified in V4L2. However,
> this is not how relative zoom is specified in UVC.
>
> UVC specifies relative zoom as a control that starts a zoom focal length
> modification at a given speed in the given direction until interrupted by
> the user (through the relative zoom control) or by a limit in the range of
> motion. This behaviour is closer to what a user would expect when
> controlling the zoom relatively : pressing a button would start zooming in
> or out, and releasing the button would stop zooming. A single V4L2 control
> encoded as a signed integer can set the speed and direction. The speed
> range should be device dependant.
>
> We are thus facing a situation where three types of zoom controls can be
> implemented, among which two are of the relative type. The UVC
> specification specifies a continuous relative zoom only, but V4L2 already
> uses the _RELATIVE suffix for incremental relative pan, tilt and focus
> controls. Using V4L2_CID_ZOOM_RELATIVE for continuous relative zoom would
> not be consistent with the V4L2 pan, tilt and focus controls, while using
> V4L2_CID_ZOOM_RELATIVE for incremental relative zoom would not be
> consistent with the UVC specification. As the V4L2 specification is already
> not consistent with the UVC specification when it comes to relative pan,
> tilt and focus, I propose to call the incremental relative zoom control
> V4L2_CID_ZOOM_RELATIVE and use a different name for the continuous control.
> Comments are welcome, as well as suggestions for the control name.
>
> Continuous relative zoom also suffers from another issue. While absolute
> and incremental relative zoom do not interact with digital zoom (when
> implemented by the device), it might be interesting to let the continuous
> relative zoom use digital zoom as an option when reaching the end of the
> optical zoom capabilities. This would give the user a large zoom range
> combining optical and digital zoom that can be navigated using a single
> control. This is how the UVC specification defines the relative zoom
> control. We would then need an additional control to enable or disable
> digital zoom when using the continuous relative zoom. Both the digital zoom
> enable and continuous relative zoom (sign + speed) values should then be
> set in a single operation through the extended controls API. Comments on
> this subject are welcome as well.

Best regards,

Laurent Pinchart

_______________________________________________
v4l-dvb-maintainer mailing list
v4l-dvb-maintainer@xxxxxxxxxxx
http://www.linuxtv.org/cgi-bin/mailman/listinfo/v4l-dvb-maintainer