[libcamera-devel] [PATCH 01/11] ipa: ipu3: Document IPAIPU3 class interface

[Laurent Pinchart]

Hi Jean-Michel,

Thank you for the patch.

On Mon, Sep 13, 2021 at 04:58:00PM +0200, Jean-Michel Hautbois wrote:
> The IPU3 IPA is maturing to a modular and extensible system capable of
> handling specific algorithms for the accelleration clusters on the ImgU.

s/accelleration/acceleration/

Where does the term "acceleration cluster" come from ? It sounds weird
to me, it's the first time I come across it for ISPs. Have I missed
something ? Or should we say "processing blocks", "processing
operations" or something similar ?

> Provide a top-level class documentation to provide an overview of the
> IPA, detailing what events are used and what algorithms are currently
> supported, as well as the limitations currently imposed.
> 
> Signed-off-by: Jean-Michel Hautbois <jeanmichel.hautbois at ideasonboard.com>
> Signed-off-by: Kieran Bingham <kieran.bingham at ideasonboard.com>
> ---
>  src/ipa/ipu3/ipu3.cpp | 64 +++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 64 insertions(+)
> 
> diff --git a/src/ipa/ipu3/ipu3.cpp b/src/ipa/ipu3/ipu3.cpp
> index c925cf64..90053069 100644
> --- a/src/ipa/ipu3/ipu3.cpp
> +++ b/src/ipa/ipu3/ipu3.cpp
> @@ -145,6 +145,70 @@ LOG_DEFINE_CATEGORY(IPAIPU3)
>  
>  namespace ipa::ipu3 {
>  
> +/**
> + * \brief The IPU3 IPA implementation
> + *
> + * The IPU3 Pipeline defines an IPU3 specific interface for communication

s/IPU3 specific/IPU3-specific/

As far as I know, a hyphen is used for compound adjectives (an
IPU3-specific interface) but not for predicates (the interface is IPU3
specific).

There are three other instances below.

> + * between the PipelineHandler, and the IPA module.

s/PipelineHandler,/PipelineHandler/

> + *
> + * We extend the IPAIPU3Interface to implement our algorithms and handle events
> + * from the IPU3 PipelineHandler to satisfy requests from the application.
> + *
> + * At initialisation time, a CameraSensorHelper is instantiated to support
> + * camera specific calculations, while the default controls are computed, and
> + * the algorithms are constructed and placed in an ordered list.
> + *
> + * The IPU3 ImgU operates with a grid layout to divide the overall frame into
> + * rectangular cells of pixels. When the IPA is configured, we determine the
> + * best grid for the statistics based on the pipeline handler Bayer Down Scaler
> + * output size.

On a side note, I wonder if the IPA should be involved to compute the
ImgU configuration. It's a topic for future research.

> + *
> + * Two main events are then handled to facilitate the operation of the IPU3 ImgU

I think we do a bit more than "facilitating" ImgU operation :-)

> + * by populating its parameter buffer, and adapting the settings of the sensor
> + * attached to the IPU3 CIO2 through sensor specific V4L2 controls.
> + *
> + * When the event \a EventFillParams occurs we populate the ImgU parameter
> + * buffer with settings to configure the device in preparation for handling the
> + * frame queued in the Request.
> + *
> + * When the frame has completed processing, the ImgU will generate a statistics
> + * buffer which is given to the IPA as part of the \a EventStatReady event. At
> + * this event we run the algorithms to parse the statistics and cache any
> + * results for the next \a EventFillParams event.
> + *
> + * The individual algorithms are split into modular components that are called
> + * iteratively to allow them to process statistics from the ImgU in a defined
> + * order.
> + *
> + * The current implementation supports three core algorithms:
> + * - Automatic white balance (AWB)
> + * - Automatic gain and exposure control (AGC)
> + * - Tonemapping (Gamma)

s/Tonemapping/Tone mapping/

Same below.

> + *
> + * AWB is implemented using a Greyworld algorithm, and calculates the red and
> + * blue gains to apply to generate a neutral grey frame overall.
> + *
> + * AGC is handled by calculating a histogram of the green channel to estimate an
> + * analogue gain and shutter time which will provide a well exposed frame. An
> + * IIR filter is used to smooth the changes to the sensor to reduce perceivable

I'd say "low-pass filter" (of possibly "low-pass IIR filter") instead of
"IIR filter" here, as the infinite impulse response isn't what really
matters.

> + * steps.
> + *
> + * The Tonemapping algorithm provides a gamma correction table to improve the
> + * contrast of the scene.
> + *
> + * The IPU3 ImgU has further accellerator clusters to support image quality

s/accellerator/accelerator/

Same comment as for the commit message.

> + * improvements through bayer and temporal noise reductions, however those are
> + * not supported in the current implementation, and will use default settings as
> + * provided by the kernel driver.
> + *
> + * Demosaicing is operating on the default values and could be further optimised

I'm not totally sure what "values" mean here, did you mean "operating
with the default parameters" ?

> + * to provide improved sharpening coefficients, checker artifact removal, and
> + * false color correction.
> + *
> + * Additional image enhancements can be made by providing lens and sensor
> + * specific tuning to adapt for Black Level compensation (BLC), Lens shading
> + * correction (SHD) and Color correction (CCM)

s/$/./

> + */
>  class IPAIPU3 : public IPAIPU3Interface
>  {
>  public:

-- 
Regards,

Laurent Pinchart