[libcamera-devel] [PATCH v3] libcamera: properties: Re-define ScalerCropMaximum

Jacopo Mondi jacopo at jmondi.org
Mon Dec 7 12:24:45 CET 2020


Hi Laurent,
  gentle ping

On Tue, Dec 01, 2020 at 06:01:57PM +0100, Jacopo Mondi wrote:
> Hi Laurent,
>
> On Tue, Dec 01, 2020 at 03:50:39PM +0200, Laurent Pinchart wrote:
> > Hi Jacopo,
> >
> > Thank you for the patch.
> >

[snip]

> > > +        Implementation details for the minimum valid crop rectangle. If a
> > > +        pipeline cannot up-scale, the minimum valid crop rectangle is equal to
> > > +        the size of sensor pixel array portion used to produce the smallest
> > > +        available stream resolution, accounting for any applied pixel
> > > +        sub-sampling technique in use and including any border pixels required
> > > +        by the ISP for image processing.
> >
> > I'm having trouble parsing this. By "smallest available stream
> > resolution", are you talking about the smallest resolution a camera can
> > produce, or the smallest stream in the current configuration ? In the
> > latter case, shouldn't it be the largest stream, as upscaling is not
> > possible ?
>
> I meant the "smallest resolution in the current configuration"
> and I think "the smallest" is right, but I get what you mean.
>
> It's not easy to express this with words but let me try.
>
>   What I meant is that the min crop rectangle should be equal to the
>   analog crop rectangle of the smaller sensor mode the pipeline can
>   select, when considered in isolation.
>
>   I thought at this a static property mostly, that will give you the
>   possibility to calculate the largest possible zoom factor, to be
>   populated at camera initialization time (to make it possible to
>   report ANDROID_SCALER_MAX_DIGITAL_ZOOM). When a mode is applied, the
>   pipeline will of course select a sensor mode large enough to
>   accommodate the largest stream configuration (assuming no
>   up-scaling) and update this. To me it was implied, but I agree it's
>   not clear.
>
> So, I guess the thing boils down to the fact I want to use this also
> to report the max digital zoom factor to android, and this is an
> absolute value, so I envisioned pipelines to initialize this at the
> smallest selectable sensor mode.
>
> How should we reword this to make it fit for both usages ?
>


More information about the libcamera-devel mailing list