path: root/src/ipa/ipu3/algorithms/agc.cpp

/* SPDX-License-Identifier: LGPL-2.1-or-later */
/*
 * Copyright (C) 2021, Ideas On Board
 *
 * ipu3_agc.cpp - AGC/AEC mean-based control algorithm
 */

#include "agc.h"

#include <algorithm>
#include <chrono>
#include <cmath>

#include <libcamera/base/log.h>
#include <libcamera/base/utils.h>

#include <libcamera/control_ids.h>
#include <libcamera/ipa/core_ipa_interface.h>

#include "libipa/histogram.h"

/**
 * \file agc.h
 */

namespace libcamera {

using namespace std::literals::chrono_literals;

namespace ipa::ipu3::algorithms {

/**
 * \class Agc
 * \brief A mean-based auto-exposure algorithm
 *
 * This algorithm calculates a shutter time and an analogue gain so that the
 * average value of the green channel of the brightest 2% of pixels approaches
 * 0.5. The AWB gains are not used here, and all cells in the grid have the same
 * weight, like an average-metering case. In this metering mode, the camera uses
 * light information from the entire scene and creates an average for the final
 * exposure setting, giving no weighting to any particular portion of the
 * metered area.
 *
 * Reference: Battiato, Messina & Castorina. (2008). Exposure
 * Correction for Imaging Devices: An Overview. 10.1201/9781420054538.ch12.
 */

LOG_DEFINE_CATEGORY(IPU3Agc)

/* Minimum limit for analogue gain value */
static constexpr double kMinAnalogueGain = 1.0;

/* \todo Honour the FrameDurationLimits control instead of hardcoding a limit */
static constexpr utils::Duration kMaxShutterSpeed = 60ms;

/* Histogram constants */
static constexpr uint32_t knumHistogramBins = 256;

/* Target value to reach for the top 2% of the histogram */
static constexpr double kEvGainTarget = 0.5;

/* Number of frames to wait before calculating stats on minimum exposure */
static constexpr uint32_t kNumStartupFrames = 10;

/*
 * Relative luminance target.
 *
 * It's a number that's chosen so that, when the camera points at a grey
 * target, the resulting image brightness is considered right.
 */
static constexpr double kRelativeLuminanceTarget = 0.16;

Agc::Agc()
	: minShutterSpeed_(0s), maxShutterSpeed_(0s)
{
}

/**
 * \brief Initialise the AGC algorithm from tuning files
 * \param[in] context The shared IPA context
 * \param[in] tuningData The YamlObject containing Agc tuning data
 *
 * This function calls the base class' tuningData parsers to discover which
 * control values are supported.
 *
 * \return 0 on success or errors from the base class
 */
int Agc::init(IPAContext &context, const YamlObject &tuningData)
{
	int ret;

	ret = parseTuningData(tuningData);
	if (ret)
		return ret;

	context.ctrlMap.merge(controls());

	return 0;
}

/**
 * \brief Configure the AGC given a configInfo
 * \param[in] context The shared IPA context
 * \param[in] configInfo The IPA configuration data
 *
 * \return 0
 */
int Agc::configure(IPAContext &context,
		   [[maybe_unused]] const IPAConfigInfo &configInfo)
{
	const IPASessionConfiguration &configuration = context.configuration;
	IPAActiveState &activeState = context.activeState;

	stride_ = configuration.grid.stride;
	bdsGrid_ = configuration.grid.bdsGrid;

	minShutterSpeed_ = configuration.agc.minShutterSpeed;
	maxShutterSpeed_ = std::min(configuration.agc.maxShutterSpeed,
				    kMaxShutterSpeed);

	minAnalogueGain_ = std::max(configuration.agc.minAnalogueGain, kMinAnalogueGain);
	maxAnalogueGain_ = configuration.agc.maxAnalogueGain;

	/* Configure the default exposure and gain. */
	activeState.agc.gain = minAnalogueGain_;
	activeState.agc.exposure = 10ms / configuration.sensor.lineDuration;

	frameCount_ = 0;

	context.activeState.agc.constraintMode = constraintModes().begin()->first;
	context.activeState.agc.exposureMode = exposureModeHelpers().begin()->first;

	/* \todo Run this again when FrameDurationLimits is passed in */
	setLimits(minShutterSpeed_, maxShutterSpeed_, minAnalogueGain_,
		  maxAnalogueGain_);
	resetFrameCount();

	return 0;
}

/**
 * \brief Estimate the mean value of the top 2% of the histogram
 * \param[in] stats The statistics computed by the ImgU
 * \param[in] grid The grid used to store the statistics in the IPU3
 * \return The mean value of the top 2% of the histogram
 */
double Agc::measureBrightness(const ipu3_uapi_stats_3a *stats,
			      const ipu3_uapi_grid_config &grid) const
{
	/* Initialise the histogram array */
	uint32_t hist[knumHistogramBins] = { 0 };

	for (unsigned int cellY = 0; cellY < grid.height; cellY++) {
		for (unsigned int cellX = 0; cellX < grid.width; cellX++) {
			uint32_t cellPosition = cellY * stride_ + cellX;

			const ipu3_uapi_awb_set_item *cell =
				reinterpret_cast<const ipu3_uapi_awb_set_item *>(
					&stats->awb_raw_buffer.meta_data[cellPosition]
				);

			uint8_t gr = cell->Gr_avg;
			uint8_t gb = cell->Gb_avg;
			/*
			 * Store the average green value to estimate the
			 * brightness. Even the overexposed pixels are
			 * taken into account.
			 */
			hist[(gr + gb) / 2]++;
		}
	}

	/* Estimate the quantile mean of the top 2% of the histogram. */
	return Histogram(Span<uint32_t>(hist)).interQuantileMean(0.98, 1.0);
}

Histogram Agc::parseStatistics(const ipu3_uapi_stats_3a *stats,
			       const ipu3_uapi_grid_config &grid)
{
	uint32_t hist[knumHistogramBins] = { 0 };

	rgbTriples_.clear();

	for (unsigned int cellY = 0; cellY < grid.height; cellY++) {
		for (unsigned int cellX = 0; cellX < grid.width; cellX++) {
			uint32_t cellPosition = cellY * stride_ + cellX;

			const ipu3_uapi_awb_set_item *cell =
				reinterpret_cast<const ipu3_uapi_awb_set_item *>(
					&stats->awb_raw_buffer.meta_data[cellPosition]);

			rgbTriples_.push_back({
				cell->R_avg,
				(cell->Gr_avg + cell->Gb_avg) / 2,
				cell->B_avg
			});

			/*
			 * Store the average green value to estimate the
			 * brightness. Even the overexposed pixels are
			 * taken into account.
			 */
			hist[(cell->Gr_avg + cell->Gb_avg) / 2]++;
		}
	}

	return Histogram(Span<uint32_t>(hist));
}

/**
 * \brief Apply a filter on the exposure value to limit the speed of changes
 * \param[in] exposureValue The target exposure from the AGC algorithm
 *
 * The speed of the filter is adaptive, and will produce the target quicker
 * during startup, or when the target exposure is within 20% of the most recent
 * filter output.
 *
 * \return The filtered exposure
 */
utils::Duration Agc::filterExposure(utils::Duration exposureValue)
{
	double speed = 0.2;

	/* Adapt instantly if we are in startup phase. */
	if (frameCount_ < kNumStartupFrames)
		speed = 1.0;

	/*
	 * If we are close to the desired result, go faster to avoid making
	 * multiple micro-adjustments.
	 * \todo Make this customisable?
	 */
	if (filteredExposure_ < 1.2 * exposureValue &&
	    filteredExposure_ > 0.8 * exposureValue)
		speed = sqrt(speed);

	filteredExposure_ = speed * exposureValue +
			    filteredExposure_ * (1.0 - speed);

	LOG(IPU3Agc, Debug) << "After filtering, exposure " << filteredExposure_;

	return filteredExposure_;
}

/**
 * \brief Estimate the new exposure and gain values
 * \param[inout] frameContext The shared IPA frame Context
 * \param[in] yGain The gain calculated based on the relative luminance target
 * \param[in] iqMeanGain The gain calculated based on the relative luminance target
 */
void Agc::computeExposure(IPAContext &context, IPAFrameContext &frameContext,
			  double yGain, double iqMeanGain)
{
	const IPASessionConfiguration &configuration = context.configuration;
	/* Get the effective exposure and gain applied on the sensor. */
	uint32_t exposure = frameContext.sensor.exposure;
	double analogueGain = frameContext.sensor.gain;

	/* Use the highest of the two gain estimates. */
	double evGain = std::max(yGain, iqMeanGain);

	/* Consider within 1% of the target as correctly exposed */
	if (utils::abs_diff(evGain, 1.0) < 0.01)
		LOG(IPU3Agc, Debug) << "We are well exposed (evGain = "
				    << evGain << ")";

	/* extracted from Rpi::Agc::computeTargetExposure */

	/* Calculate the shutter time in seconds */
	utils::Duration currentShutter = exposure * configuration.sensor.lineDuration;

	/*
	 * Update the exposure value for the next computation using the values
	 * of exposure and gain really used by the sensor.
	 */
	utils::Duration effectiveExposureValue = currentShutter * analogueGain;

	LOG(IPU3Agc, Debug) << "Actual total exposure " << currentShutter * analogueGain
			    << " Shutter speed " << currentShutter
			    << " Gain " << analogueGain
			    << " Needed ev gain " << evGain;

	/*
	 * Calculate the current exposure value for the scene as the latest
	 * exposure value applied multiplied by the new estimated gain.
	 */
	utils::Duration exposureValue = effectiveExposureValue * evGain;

	/* Clamp the exposure value to the min and max authorized */
	utils::Duration maxTotalExposure = maxShutterSpeed_ * maxAnalogueGain_;
	exposureValue = std::min(exposureValue, maxTotalExposure);
	LOG(IPU3Agc, Debug) << "Target total exposure " << exposureValue
			    << ", maximum is " << maxTotalExposure;

	/*
	 * Filter the exposure.
	 * \todo estimate if we need to desaturate
	 */
	exposureValue = filterExposure(exposureValue);

	/*
	 * Divide the exposure value as new exposure and gain values.
	 *
	 * Push the shutter time up to the maximum first, and only then
	 * increase the gain.
	 */
	utils::Duration shutterTime =
		std::clamp<utils::Duration>(exposureValue / minAnalogueGain_,
					    minShutterSpeed_, maxShutterSpeed_);
	double stepGain = std::clamp(exposureValue / shutterTime,
				     minAnalogueGain_, maxAnalogueGain_);
	LOG(IPU3Agc, Debug) << "Divided up shutter and gain are "
			    << shutterTime << " and "
			    << stepGain;
}

/**
 * \brief Estimate the relative luminance of the frame with a given gain
 * \param[in] frameContext The shared IPA frame context
 * \param[in] grid The grid used to store the statistics in the IPU3
 * \param[in] stats The IPU3 statistics and ISP results
 * \param[in] gain The gain to apply to the frame
 * \return The relative luminance
 *
 * This function estimates the average relative luminance of the frame that
 * would be output by the sensor if an additional \a gain was applied.
 *
 * The estimation is based on the AWB statistics for the current frame. Red,
 * green and blue averages for all cells are first multiplied by the gain, and
 * then saturated to approximate the sensor behaviour at high brightness
 * values. The approximation is quite rough, as it doesn't take into account
 * non-linearities when approaching saturation.
 *
 * The relative luminance (Y) is computed from the linear RGB components using
 * the Rec. 601 formula. The values are normalized to the [0.0, 1.0] range,
 * where 1.0 corresponds to a theoretical perfect reflector of 100% reference
 * white.
 *
 * More detailed information can be found in:
 * https://en.wikipedia.org/wiki/Relative_luminance
 */
double Agc::estimateLuminance(IPAActiveState &activeState,
			      const ipu3_uapi_grid_config &grid,
			      const ipu3_uapi_stats_3a *stats,
			      double gain)
{
	double redSum = 0, greenSum = 0, blueSum = 0;

	/* Sum the per-channel averages, saturated to 255. */
	for (unsigned int cellY = 0; cellY < grid.height; cellY++) {
		for (unsigned int cellX = 0; cellX < grid.width; cellX++) {
			uint32_t cellPosition = cellY * stride_ + cellX;

			const ipu3_uapi_awb_set_item *cell =
				reinterpret_cast<const ipu3_uapi_awb_set_item *>(
					&stats->awb_raw_buffer.meta_data[cellPosition]
				);
			const uint8_t G_avg = (cell->Gr_avg + cell->Gb_avg) / 2;

			redSum += std::min(cell->R_avg * gain, 255.0);
			greenSum += std::min(G_avg * gain, 255.0);
			blueSum += std::min(cell->B_avg * gain, 255.0);
		}
	}

	/*
	 * Apply the AWB gains to approximate colours correctly, use the Rec.
	 * 601 formula to calculate the relative luminance, and normalize it.
	 */
	double ySum = redSum * activeState.awb.gains.red * 0.299
		    + greenSum * activeState.awb.gains.green * 0.587
		    + blueSum * activeState.awb.gains.blue * 0.114;

	return ySum / (grid.height * grid.width) / 255;
}

double Agc::estimateLuminance(double gain) const
{
	double redSum = 0, greenSum = 0, blueSum = 0;

	for (unsigned int i = 0; i < rgbTriples_.size(); i++) {
		redSum += std::min(std::get<0>(rgbTriples_[i]) * gain, 255.0);
		greenSum += std::min(std::get<1>(rgbTriples_[i]) * gain, 255.0);
		blueSum += std::min(std::get<2>(rgbTriples_[i]) * gain, 255.0);
	}

	double ySum = redSum * rGain_ * 0.299
		    + greenSum * gGain_ * 0.587
		    + blueSum * bGain_ * 0.114;

	return ySum / (bdsGrid_.height * bdsGrid_.width) / 255;
}

/**
 * \brief Process IPU3 statistics, and run AGC operations
 * \param[in] context The shared IPA context
 * \param[in] frame The current frame sequence number
 * \param[in] frameContext The current frame context
 * \param[in] stats The IPU3 statistics and ISP results
 * \param[out] metadata Metadata for the frame, to be filled by the algorithm
 *
 * Identify the current image brightness, and use that to estimate the optimal
 * new exposure and gain for the scene.
 */
void Agc::process(IPAContext &context, [[maybe_unused]] const uint32_t frame,
		  IPAFrameContext &frameContext,
		  const ipu3_uapi_stats_3a *stats,
		  ControlList &metadata)
{
	/*
	 * Estimate the gain needed to have the proportion of pixels in a given
	 * desired range. iqMean is the mean value of the top 2% of the
	 * cumulative histogram, and we want it to be as close as possible to a
	 * configured target.
	 */
	double iqMean = measureBrightness(stats, context.configuration.grid.bdsGrid);
	double iqMeanGain = kEvGainTarget * knumHistogramBins / iqMean;

	/*
	 * Estimate the gain needed to achieve a relative luminance target. To
	 * account for non-linearity caused by saturation, the value needs to be
	 * estimated in an iterative process, as multiplying by a gain will not
	 * increase the relative luminance by the same factor if some image
	 * regions are saturated.
	 */
	double yGain = 1.0;
	double yTarget = kRelativeLuminanceTarget;

	for (unsigned int i = 0; i < 8; i++) {
		double yValue = estimateLuminance(context.activeState,
						  context.configuration.grid.bdsGrid,
						  stats, yGain);
		double extraGain = std::min(10.0, yTarget / (yValue + .001));

		yGain *= extraGain;
		LOG(IPU3Agc, Debug) << "Y value: " << yValue
				    << ", Y target: " << yTarget
				    << ", gives gain " << yGain;
		if (extraGain < 1.01)
			break;
	}

	computeExposure(context, frameContext, yGain, iqMeanGain);
	frameCount_++;

	Histogram hist = parseStatistics(stats, context.configuration.grid.bdsGrid);
	rGain_ = context.activeState.awb.gains.red;
	gGain_ = context.activeState.awb.gains.blue;
	bGain_ = context.activeState.awb.gains.green;

	/*
	 * The Agc algorithm needs to know the effective exposure value that was
	 * applied to the sensor when the statistics were collected.
	 */
	utils::Duration exposureTime = context.configuration.sensor.lineDuration
				     * frameContext.sensor.exposure;
	double analogueGain = frameContext.sensor.gain;
	utils::Duration effectiveExposureValue = exposureTime * analogueGain;

	utils::Duration shutterTime;
	double aGain, dGain;
	std::tie(shutterTime, aGain, dGain) =
		calculateNewEv(context.activeState.agc.constraintMode,
			       context.activeState.agc.exposureMode, hist,
			       effectiveExposureValue);

	LOG(IPU3Agc, Debug)
		<< "Divided up shutter, analogue gain and digital gain are "
		<< shutterTime << ", " << aGain << " and " << dGain;

	IPAActiveState &activeState = context.activeState;
	/* Update the estimated exposure and gain. */
	activeState.agc.exposure = shutterTime / context.configuration.sensor.lineDuration;
	activeState.agc.gain = aGain;

	metadata.set(controls::AnalogueGain, frameContext.sensor.gain);
	metadata.set(controls::ExposureTime, exposureTime.get<std::micro>());

	/* \todo Use VBlank value calculated from each frame exposure. */
	uint32_t vTotal = context.configuration.sensor.size.height
			+ context.configuration.sensor.defVBlank;
	utils::Duration frameDuration = context.configuration.sensor.lineDuration
				      * vTotal;
	metadata.set(controls::FrameDuration, frameDuration.get<std::micro>());

}

REGISTER_IPA_ALGORITHM(Agc, "Agc")

} /* namespace ipa::ipu3::algorithms */

} /* namespace libcamera */