[libcamera-devel] [PATCH v2 2/4] libcamera: camera: Introduce SensorConfiguration

[Jacopo Mondi]

Hi Kieran

On Thu, Sep 14, 2023 at 11:25:18PM +0100, Kieran Bingham via libcamera-devel wrote:
> Quoting Jacopo Mondi via libcamera-devel (2023-07-31 12:31:13)
> > Introduce SensorConfiguration in the libcamera API.
> >
> > The SensorConfiguration is part of the CameraConfiguration class
> > and allows applications to control the sensor settings.
> >
> > Signed-off-by: Jacopo Mondi <jacopo.mondi at ideasonboard.com>
> > Reviewed-by: Naushir Patuck <naush at raspberrypi.com>
> > ---
> >  include/libcamera/camera.h |  43 +++++++++
> >  src/libcamera/camera.cpp   | 180 +++++++++++++++++++++++++++++++++++++
> >  2 files changed, 223 insertions(+)
> >
> > diff --git a/include/libcamera/camera.h b/include/libcamera/camera.h
> > index 004bc89455f5..b2aa8d467feb 100644
> > --- a/include/libcamera/camera.h
> > +++ b/include/libcamera/camera.h
> > @@ -19,6 +19,7 @@
> >  #include <libcamera/base/signal.h>
> >
> >  #include <libcamera/controls.h>
> > +#include <libcamera/geometry.h>
> >  #include <libcamera/request.h>
> >  #include <libcamera/stream.h>
> >  #include <libcamera/transform.h>
> > @@ -30,6 +31,47 @@ class FrameBufferAllocator;
> >  class PipelineHandler;
> >  class Request;
> >
> > +class SensorConfiguration
> > +{
> > +public:
> > +       unsigned int bitDepth = 0;
> > +
> > +       Rectangle analogCrop;
> > +
> > +       struct {
> > +               unsigned int binX = 1;
> > +               unsigned int binY = 1;
> > +       } binning;
> > +
> > +       struct {
> > +               unsigned int xOddInc = 1;
> > +               unsigned int xEvenInc = 1;
> > +               unsigned int yOddInc = 1;
> > +               unsigned int yEvenInc = 1;
> > +       } skipping;
> > +
> > +       Size outputSize;
> > +
> > +       bool valid() const
> > +       {
> > +               return validate() != Invalid;
> > +       }
> > +
> > +       explicit operator bool() const
> > +       {
> > +               return validate() == Populated;
> > +       }
> > +
> > +private:
> > +       enum Status {
> > +               Unpopulated,
> > +               Populated,
> > +               Invalid,
> > +       };
> > +
> > +       Status validate() const;
> > +};
> > +
> >  class CameraConfiguration
> >  {
> >  public:
> > @@ -66,6 +108,7 @@ public:
> >         bool empty() const;
> >         std::size_t size() const;
> >
> > +       SensorConfiguration sensorConfig;
> >         Transform transform;
> >
> >  protected:
> > diff --git a/src/libcamera/camera.cpp b/src/libcamera/camera.cpp
> > index 0eecee766f00..e796a5da28c6 100644
> > --- a/src/libcamera/camera.cpp
> > +++ b/src/libcamera/camera.cpp
> > @@ -97,6 +97,21 @@
> >   * implemented in the above order at the hardware level. The libcamera pipeline
> >   * handlers translate the pipeline model to the real hardware configuration.
> >   *
> > + * \subsection camera-sensor-mode Camera Sensor Model
> > + *
> > + * libcamera allows applications to control the configuration of the camera
> > + * sensor, particularly it allows to control the frame geometry and frame
> > + * delivery rate of the sensor.
> > + *
> > + * The camera sensor configuration applies to all streams produced by a camera
> > + * from the same image source.
> > + *
> > + * More details about the libcamera implemented camera sensor model are
>
> More details about the camera sensor model implemented by libcamera are
>
> > + * available in the libcamera camera-sensor-model documentation page.
> > + *
> > + * The sensor configuration is specified by applications by populating the
> > + * CameraConfiguration::sensorConfig class member.
> > + *
> >   * \subsection digital-zoom Digital Zoom
> >   *
> >   * Digital zoom is implemented as a combination of the cropping and scaling
> > @@ -111,6 +126,161 @@ namespace libcamera {
> >
> >  LOG_DECLARE_CATEGORY(Camera)
> >
> > +/**
> > + * \class SensorConfiguration
> > + * \brief Camera sensor configuration
> > + *
> > + * The SensorConfiguration class collects parameters to control the operations
> > + * of the camera sensor, accordingly to the abstract camera sensor model
> > + * implemented by libcamera.
> > + *
> > + * \todo Applications shall fully populate all fields of the
> > + * CameraConfiguration::sensorConfig class members before validating the
> > + * CameraConfiguration. If the SensorConfiguration is not fully populated, or if
> > + * any of its parameters cannot be applied to the sensor in use, the
> > + * CameraConfiguration validation process will fail and return
> > + * CameraConfiguration::Status::Invalid.
> > + *
> > + * Applications that populate the SensorConfiguration class members are
> > + * expected to be highly-specialized applications that know what sensor
> > + * they are operating with and what parameters are valid for the sensor in use.
> > + *
> > + * A detailed description of the abstract camera sensor model implemented by
> > + * libcamera and the description of its configuration parameters is available
> > + * in the libcamera documentation camera-sensor-model file.
> > + */
> > +
> > +/**
> > + * \enum SensorConfiguration::Status
> > + * \brief The sensor configuration validation status
> > + */
> > +
> > +/**
> > + * \var SensorConfiguration::bitDepth
> > + * \brief The sensor image format bit depth
> > + *
> > + * The number of bits (resolution) used to represent a pixel sample.
> > + */
> > +
> > +/**
> > + * \var SensorConfiguration::analogCrop
> > + * \brief The analog crop rectangle
> > + *
> > + * The selected portion of the active pixel array used to produce the image
> > + * frame.
> > + */
> > +
> > +/**
> > + * \var SensorConfiguration::binning
> > + * \brief Sensor binning configuration
> > + *
> > + * Refer to the camera-sensor-model documentation for an accurate description
> > + * of the binning operations. Disabled by default.
> > + */
> > +
> > +/**
> > + * \var SensorConfiguration::binX
> > + * \brief Horizontal binning factor
> > + *
> > + * The horizontal binning factor. Default to 1.
> > + */
> > +
> > +/**
> > + * \var SensorConfiguration::binY
> > + * \brief Vertical binning factor
> > + *
> > + * The vertical binning factor. Default to 1.
> > + */
> > +
> > +/**
> > + * \var SensorConfiguration::skipping
> > + * \brief The sensor skipping configuration
> > + *
> > + * Refer to the camera-sensor-model documentation for an accurate description
> > + * of the skipping operations.
> > + *
> > + * If no skipping is performed, all the structure fields should be
> > + * set to 1. Disabled by default.
> > + */
> > +
> > +/**
> > + * \var SensorConfiguration::xOddInc
> > + * \brief Horizontal increment for odd rows. Default to 1.
> > + */
> > +
> > +/**
> > + * \var SensorConfiguration::xEvenInc
> > + * \brief Horizontal increment for even rows. Default to 1.
> > + */
> > +
> > +/**
> > + * \var SensorConfiguration::yOddInc
> > + * \brief Vertical increment for odd columns. Default to 1.
> > + */
> > +
> > +/**
> > + * \var SensorConfiguration::yEvenInc
> > + * \brief Vertical increment for even columns. Default to 1.
> > + */
> > +
> > +/**
> > + * \var SensorConfiguration::outputSize
> > + * \brief The frame output (visible) size
> > + *
> > + * The size of the data frame as received by the host processor.
> > + */
> > +
> > +/**
> > + * \fn SensorConfiguration::valid() const
> > + * \brief Validate the SensorConfiguration
> > + *
> > + * Validate the sensor configuration.
> > + *
> > + * \todo A sensor configuration is valid (or well-formed) if it's either
> > + * completely un-populated or fully populated. For now allow applications to
> > + * populate the bitDepth and the outputSize only.
> > + *
> > + * \return True if the SensorConfiguration is either fully populated or
> > + * un-populated, false otherwise
> > + */
> > +
> > +/**
> > + * \fn SensorConfiguration::operator bool() const
> > + * \brief Test if a SensorConfiguration is fully populated
> > + * \return True if the SensorConfiguration is fully populated
> > + */
> > +
> > +/**
> > + * \brief Validate the sensor configuration
> > + *
> > + * \todo A sensor configuration is valid (or well-formed) if it's either
> > + * completely un-populated or fully populated. For now allow applications to
> > + * populate the bitDepth and the outputSize only.
> > + *
> > + * \return The sensor configuration status
> > + * \retval Unpopulated The sensor configuration is fully unpopulated
> > + * \retval Populated The sensor configuration is fully populated
> > + * \retval Invalid The sensor configuration is invalid (not fully populated
> > + * and not fully unpopulated)
> > + */
> > +SensorConfiguration::Status SensorConfiguration::validate() const
> > +{
> > +       if (bitDepth && binning.binX && binning.binY &&
> > +           skipping.xOddInc && skipping.yOddInc &&
> > +           skipping.xEvenInc && skipping.yEvenInc &&
> > +           !outputSize.isNull())
> > +               return Populated;
> > +
> > +       if (!bitDepth &&
> > +           binning.binX <= 1 && binning.binY <= 1 &&
> > +           skipping.xOddInc <= 1 && skipping.yOddInc <= 1 &&
> > +           skipping.xEvenInc <= 1 && skipping.yEvenInc <= 1 &&
>
> This looks like the <= is to support if the struct / class has been
> unintialised/memset or somehow allocated with
>
> 	SensorConfiguration = {}; ?
>
> Otherwise - is the <= necessary ?
>
> I don't mind keeping it - but a short comment above could explain why we
> might accept skipping.yOddInc == 0 as Unpopulated rather than invalid.
>

To be honest, this has gone through a few changes, the latest one was
to default skipping and binning to 1, while they were previously
initialized to 0. That's probably where the <= comes from.

However I think it's correct, as zero-initialized SensorConfiguration
should be considered unpopulated, so I can add a a comment like

        /*
	 * By default the binning and skipping factors are initialized to 1, but
	 * a zero-initialized SensorConfiguration is considered unpopulated
	 * as well.
	 */

>
> > +           outputSize.isNull())
> > +               return Unpopulated;
> > +
> > +       return Invalid;
> > +}
> > +
> >  /**
> >   * \class CameraConfiguration
> >   * \brief Hold configuration for streams of the camera
> > @@ -391,6 +561,16 @@ CameraConfiguration::Status CameraConfiguration::validateColorSpaces(ColorSpaceF
> >         return status;
> >  }
> >
> > +/**
> > + * \var CameraConfiguration::sensorConfig
> > + * \brief The camera sensor configuration
> > + *
> > + * The sensorConfig field allows to control the configuration of the camera
>
> s/ allows to control the / allows control of the /
>
>
> But again, most of all that is easily fixed up so then:
>
>
> Reviewed-by: Kieran Bingham <kieran.bingham at ideasonboard.com>
>

Thanks
  j

> > + * sensor. Refer to the camera-sensor-model documentation and to the
> > + * SensorConfiguration class documentation for details about the sensor
> > + * configuration process.
> > + */
> > +
> >  /**
> >   * \var CameraConfiguration::transform
> >   * \brief User-specified transform to be applied to the image
> > --
> > 2.40.1
> >