[PATCH v2 8/9] libipa: agc_mean_luminance: Add exposure compensation support

[Laurent Pinchart]

On Fri, Apr 11, 2025 at 02:36:36PM +0200, Stefan Klug wrote:
> Exposure compensation allows to over- or under-expose an
> image by a given value. Add support for that in agc_mean_luminance.
> 
> The added exposure compensation can lead to luminance target values that
> are close or above saturation and are therefore never reachable. Add a
> fix for that by limiting the maximum luminance target to 0.95.
> 
> Signed-off-by: Stefan Klug <stefan.klug at ideasonboard.com>
> Reviewed-by: Daniel Scally <dan.scally at ideasonboard.com>
> 
> ---
> 
> Changes in v2:
> - Fixed compiler error that slipped through in v1
> - Improved commit message to explain the luminance target limitation
> - Collected tag
> ---
>  src/ipa/libipa/agc_mean_luminance.cpp | 23 +++++++++++++++++++++--
>  src/ipa/libipa/agc_mean_luminance.h   |  6 ++++++
>  2 files changed, 27 insertions(+), 2 deletions(-)
> 
> diff --git a/src/ipa/libipa/agc_mean_luminance.cpp b/src/ipa/libipa/agc_mean_luminance.cpp
> index 9154f083a510..f6cb7144b31a 100644
> --- a/src/ipa/libipa/agc_mean_luminance.cpp
> +++ b/src/ipa/libipa/agc_mean_luminance.cpp
> @@ -44,6 +44,15 @@ static constexpr uint32_t kNumStartupFrames = 10;
>   */
>  static constexpr double kDefaultRelativeLuminanceTarget = 0.16;
>  
> +/*
> + * Maximum relative luminance target
> + *
> + * This value limits the relative luminance target after applying the exposure
> + * compensation. Targeting a value above this limit results in saturation
> + * and the inability to regulate properly.
> + */
> +static constexpr double kMaxRelativeLuminanceTarget = 0.95;
> +
>  /**
>   * \struct AgcMeanLuminance::AgcConstraint
>   * \brief The boundaries and target for an AeConstraintMode constraint
> @@ -134,7 +143,7 @@ static constexpr double kDefaultRelativeLuminanceTarget = 0.16;
>   */
>  
>  AgcMeanLuminance::AgcMeanLuminance()
> -	: frameCount_(0), filteredExposure_(0s), relativeLuminanceTarget_(0)
> +	: exposureCompensation_(1.0), frameCount_(0), filteredExposure_(0s), relativeLuminanceTarget_(0)

Line wrap please.

>  {
>  }
>  
> @@ -369,6 +378,15 @@ int AgcMeanLuminance::parseTuningData(const YamlObject &tuningData)
>  	return parseExposureModes(tuningData);
>  }
>  
> +/**
> + * \fn AgcMeanLuminance::setExposureCompensation()
> + * \brief Set the exposure compensation value
> + * \param[in] gain The exposure compensation gain
> + *
> + * This function sets the exposure compensation value to be used in the
> + * AGC calculations. It is expressed as gain instead of EV.

What's the rationale for expressing it as a gain instead of EV ?

> + */
> +
>  /**
>   * \brief Set the ExposureModeHelper limits for this class
>   * \param[in] minExposureTime Minimum exposure time to allow
> @@ -425,7 +443,8 @@ void AgcMeanLuminance::setLimits(utils::Duration minExposureTime,
>   */
>  double AgcMeanLuminance::estimateInitialGain() const
>  {
> -	double yTarget = relativeLuminanceTarget_;
> +	double yTarget = std::min(relativeLuminanceTarget_ * exposureCompensation_,
> +				  kMaxRelativeLuminanceTarget);
>  	double yGain = 1.0;
>  
>  	/*
> diff --git a/src/ipa/libipa/agc_mean_luminance.h b/src/ipa/libipa/agc_mean_luminance.h
> index c41391cb0b73..cad7ef845487 100644
> --- a/src/ipa/libipa/agc_mean_luminance.h
> +++ b/src/ipa/libipa/agc_mean_luminance.h
> @@ -44,6 +44,11 @@ public:
>  
>  	int parseTuningData(const YamlObject &tuningData);
>  
> +	void setExposureCompensation(double gain)
> +	{
> +		exposureCompensation_ = gain;
> +	}
> +
>  	void setLimits(utils::Duration minExposureTime, utils::Duration maxExposureTime,
>  		       double minGain, double maxGain);
>  
> @@ -84,6 +89,7 @@ private:
>  				   double gain);
>  	utils::Duration filterExposure(utils::Duration exposureValue);
>  
> +	double exposureCompensation_;
>  	uint64_t frameCount_;
>  	utils::Duration filteredExposure_;
>  	double relativeLuminanceTarget_;

-- 
Regards,

Laurent Pinchart