Replace the dropFrameCount parameter returned from ipa::start() to the
pipeline handler by startupFrameCount and invalidFrameCount. The former
counts the number of frames required for AWB/AGC to converge, and the
latter counts the number of invalid frames produced by the sensor when
starting up.
In the pipeline handler, use the sum of these 2 values to replicate the
existing dropFrameCount behaviour.
Signed-off-by: Naushir Patuck <naush@raspberrypi.com>
Reviewed-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Saturation control is added on top of the colour correction matrix. A
method of saturation adjustment that can be fully integrated into the
colour correction matrix is used. The control is available only if Ccm
algorithm is enabled.
The control uses 0.0-2.0 value range, with 1.0 being unmodified
saturation, 0.0 full desaturation and 2.0 quite saturated.
The saturation is adjusted by converting to Y'CbCr colour space,
applying the saturation value on the colour axes, and converting back to
RGB. ITU-R BT.601 conversion is used to convert between the colour
spaces, for no particular reason.
The colour correction matrix is applied before gamma and the given
matrix is suitable for such a case. Alternatively, the transformation
used in libcamera rpi ccm.cpp could be used.
Signed-off-by: Milan Zamazal <mzamazal@redhat.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Even without AGC definition in the tuning file, the application would
still dereference agc unconditionally, leading to a segmentation fault
if AGC is absent.
This is relevant for sensors already providing AGC/AEC by themselves.
Check if AGC is present prior to setting maximum exposure time.
Signed-off-by: Benjamin Mugnier <benjamin.mugnier@foss.st.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Tested-by: Barnabás Pőcze <barnabas.pocze@ideasonboard.com> # RPi4 + imx708_wide
Reviewed-by: Naushir Patuck <naush@raspberrypi.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
The AWB measurements are taken after the CCM. This can be seen by
enabling debug logging on AWB, disabling AWB (stats will still be
processed) and manually chaning the CCM.
This means that the estimated colour temperature and the corresponding
CCM also lead to changed rgbMeans which in turn leads to oscillations.
Fix that by applying the inverse transform on the rgbMeans.
Signed-off-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
In the grey world AWB case, if no colour gains are contained in the
tuning file, the colour gains get reset to 1 when the colour temperature
is set manually. This is unexpected and undesirable. Allow the
gainsFromColourTemp() function to return a std::nullopt to handle that
case.
While at it, remove an unnecessary import from rkisp1/algorithms/awb.h.
Signed-off-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Add a manual ColourCorrectionMatrix control. This was already discussed
while implementing manual colour temperature but was never implemented.
The control allows to manually specify the CCM when AwbEnable is false.
Signed-off-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
In RkISP1Awb::process(), the color temperature in the active state is
updated every time new statistics are available. The CCM/LSC algorithms
use that value in prepare() to update the CCM/LSC. This is not correct
if the color temperature was specified manually and leads to visible
flicker even when AwbEnable is set to false.
To fix that, track the auto and manual color temperature separately in
active state. In Awb::prepare() the current frame context is updated
with the corresponding value from active state. Change the algorithms to
fetch the color temperature from the frame context instead of the active
state in prepare().
Fixes: 0230880954 ("ipa: rkisp1: awb: Implement ColourTemperature control")
Signed-off-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Swap gains and automatic/manual in the IPAActiveState structure. This is
in preparation to adding another member, which is easier in the new
structure. The patch contains no functional changes.
Signed-off-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
When the AWB engine doesn't find a valid pixel because all pixels lie
outside the configured colour range it returns an AWB measurement value
of 255, 255, 255. This leaves the regulation in an unrecoverable state
noticeable by a completely green image. Fix that by skipping the AWB
calculation in case there were no valid pixels.
Signed-off-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
The ControlInfo information for AwbEnable and ColourGains are declared
and exposed in the top-level IPA. These should instead be exposed by the
AWB part of the IPA, as it doesn't make sense to support these controls
when AWB is disabled, for example.
Move the declaration of these controls out of the top-level IPA and into
AWB.
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The "fast" parameter has not been used since it first appeared in the
source code. And not only is it not used, but its retrieval from
the configuration since c1597f9896 ("ipa: raspberrypi: Use YamlParser
to replace dependency on boost") has been incorrect. So remove it.
Signed-off-by: Barnabás Pőcze <barnabas.pocze@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: David Plowman <david.plowman@raspberrypi.com>
The AeEnable control is now just a wrapper that is converted to
ExposureTimeMode and AnalogueGainMode controls instead. Therefore, it
should simply be ignored when we encounter it, without the need for
any warnings.
Signed-off-by: David Plowman <david.plowman@raspberrypi.com>
Reviewed-by: Naushir Patuck <naush@raspberrypi.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
A colour correction matrix (CCM) is applied like this to an RGB pixel
vector P:
CCM * P
White balance must be applied before CCM. If CCM is used, software ISP
makes a combined matrix by multiplying the CCM by a white balance gains
CCM-like matrix (WBG). The multiplication should be as follows to do it
in the correct order:
CCM * (WBG * P) = (CCM * WBG) * P
The multiplication order in Lut software ISP algorithm is reversed,
resulting in colour casts. Let's fix the order.
Signed-off-by: Milan Zamazal <mzamazal@redhat.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Tested-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
The interQuantileMean() is supposed to return a weighted mean value
between two quantiles. This works for fine histograms, but fails for
coarse histograms and small quantile ranges because the weight is always
taken from the lower border of the bin.
Fix that by rewriting the algorithm to calculate a lower and upper bound
for every (partial) bin that goes into the mean calculation and weight
the bins by the middle of these bounds.
Signed-off-by: Stefan Klug <stefan.klug@ideasonboard.com>
Acked-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The calculation of the frac variable is based solely on integers and
therefore results in the fractional part being either 0 or 1.
In the original code from RaspberryPi this is mitigated by casting the
nominator to a double. This works for most cases, but fails when q is
very small because of the quantization introduced by item being an
integer.
Fix both issues by doing the full calculation in double and remove the
should_fail tag.
Signed-off-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
ccmEnabled variable is not initialized by default, which results in
usage of CCM when the algorithm itself is not enabled and configured.
The bug manifests itself as seldom reproducible corrupted video stream.
Fix by initialize ccmEnabled member where it is declared.
Reviewed-by: Milan Zamazal <mzamazal@redhat.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Fixes: ac30686556 ("libcamera: software_isp: Track whether CCM is enabled")
Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Automatic black level setting in software ISP updates the determined
black level value when exposure or gain change. It stores the last
exposure and gain values to detect the change.
BlackLevel::configure() resets the stored black level value but not the
exposure and gain values. This can prevent updating the black value and
cause bad image output, e.g. after suspending and resuming a camera, if
exposure and gain remain unchanged.
Let's store exposure and gain in IPAActiveState. Although the values
are not supposed to be used outside BlackLevel class, storing them in
the context has the advantage of their automatic reset together with the
other context contents and having them in `blc' struct indicates their
relationship to the black value computation.
Bug: https://bugs.libcamera.org/show_bug.cgi?id=259
Signed-off-by: Milan Zamazal <mzamazal@redhat.com>
Tested-by: Robert Mader <robert.mader@collabora.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Report exposure and gain in metadata.
This is more complicated than it could be expected because the exposure
value should be in microseconds but it's handled using V4L2_CID_EXPOSURE
control, which doesn't specify the unit, see
https://www.kernel.org/doc/html/latest/userspace-api/media/v4l/control.html.
So the unit conversion is done in the way rkisp1 IPA uses.
This requires getting and passing IPACameraSensorInfo around. To avoid
naming confusion and to improve consistency with rkisp1 IPA,
sensorCtrlInfoMap parameter is renamed to sensorControls.
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Milan Zamazal <mzamazal@redhat.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Extend the Simple IPA IPC to support returning a metadata ControlList
when the process call has completed.
A new signal from the IPA is introduced to report the metadata,
similarly to what the hardware pipelines do.
Merge the metadata reported by the ISP into any completing request to
provide to the application. Completion of a request is delayed until
this is done; this doesn't apply to canceled requests.
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Milan Zamazal <mzamazal@redhat.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
The call to setControls(0) is counter productive. At start() time, no
requests were queued and no stats were received. So setControls(0)
accesses a zeroed frame context and in turn sends 0 as gain, exposure
and vblank to the pipeline handler and DelayedControls. This leads to
strong oscillations on every start of the camera.
A proper fix for handling the startup controls still needs to be done
and was already started in [1] and [2].
From a DelayedControls point of view the call to setControls(0) is also
unnecessary as DelayedControls treat frame 0 as already being queued in
after initialization.
So it is safe to just remove it and the removal fixes the zero
effectiveExposureValue discussed in the previous patch for rkisp1.
[1]: https://patchwork.libcamera.org/patch/21708/
[2]: https://patchwork.libcamera.org/patch/22445/
Signed-off-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
In a proper system it never happens that the effectiveExposureValue
drops to zero. If that still happens due to a bug outside of
agc_mean_luminance, the calculated gain goes towards infinity but the
newExposureValue is still 0 because it is the result of multiplying the
effectiveExposureTime with the gain, leading to wild oscillations.
Catch that condition, print an error message and set the new effective
exposure value to an arbitrary 10ms.
Note that in any case the underlying problem must be fixed. The
important change is the added error message to be able to detect such a
situation.
Signed-off-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
With the availability of metering modes and the corresponding weights,
there is a flexible way of defining the area that gets taken into
account when AEGC is calculated. There is no need to reduce that window
to an arbitrary region anymore. If need arises we can make this
parameter user configurable or add a control for it.
Signed-off-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
The weights for a given metering mode are applied to the histogram data
inside the histogram statistics block. The AE statistics do not contain
any weights. Therefore the weights are honored when AgcMeanLuminance
calculates the upper or lower constraints, but ignored in the
calculation of the frame luminance. Fix that by manually applying the
weights in the luminance calculation.
Fixes: 4c5152843a ("ipa: rkisp1: Derive rkisp1::algorithms::Agc from AgcMeanLuminance")
Signed-off-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
This patch applies color correction matrix (CCM) in debayering if the
CCM is specified. Not using CCM must still be supported for performance
reasons.
The CCM is applied as follows:
[r1 g1 b1] [r]
[r2 g2 b2] * [g]
[r3 g3 b3] [b]
The CCM matrix (the left side of the multiplication) is constant during
single frame processing, while the input pixel (the right side) changes.
Because each of the color channels is only 8-bit in software ISP, we can
make 9 lookup tables with 256 input values for multiplications of each
of the r_i, g_i, b_i values. This way we don't have to multiply each
pixel, we can use table lookups and additions instead. Gamma (which is
non-linear and thus cannot be a part of the 9 lookup tables values) is
applied on the final values rounded to integers using another lookup
table.
Because the changing part is the pixel value with three color elements,
only three dynamic table lookups are needed. We use three lookup tables
to represent the multiplied matrix values, each of the tables
corresponding to the given matrix column and pixel color.
We use int16_t to store the precomputed multiplications. This seems to
be noticeably (>10%) faster than `float' for the price of slightly less
accuracy and it covers the range of values that sane CCMs produce. The
selection and structure of data is performance critical, for example
using bytes would add significant (>10%) speedup but would be too short
to cover the value range.
The color lookup tables can be represented either as unions,
accommodating tables for both the CCM and non-CCM cases, or as separate
tables for each of the cases, leaving the tables for the other case
unused. The latter is selected as a matter of preference.
The tables are copied (as before), which is not elegant but also not a
big problem. There are patches posted that use shared buffers for
parameters passing in software ISP (see software ISP TODO #5) and they
can be adjusted for the new parameter format.
Color gains from white balance are supposed not to be a part of the
specified CCM. They are applied on it using matrix multiplication,
which is simple and in correspondence with future additions in the form
of matrix multiplication, like saturation adjustment.
With this patch, the reported per-frame slowdown when applying CCM is
about 45% on Debix Model A and about 75% on TI AM69 SK.
Using std::clamp in debayering adds some performance penalty (a few
percent). The clamping is necessary to eliminate out of range values
possibly produced by the CCM. If it could be avoided by adjusting the
precomputed tables some way then performance could be improved a bit.
Signed-off-by: Milan Zamazal <mzamazal@redhat.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Applying color correction matrix (CCM) in software ISP is optional due
to performance reasons. CCM is applied if and only if `Ccm' algorithm
is present in the tuning file.
Software ISP debayering is a performance critical piece of code and we
do not want to use dynamic conditionals there. Therefore we pass
information about CCM application to debayering configuration and let it
select the right versions of debayering functions using templates. This
is a trick similar to the previously used one for adding or not adding
an alpha channel to the output.
Debayering gets this information but it ignores it in this patch.
Actual processing with CCM is added in the followup patch.
Signed-off-by: Milan Zamazal <mzamazal@redhat.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Milan Zamazal <mzamazal@redhat.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
This patch adds color correction matrix (CCM) algorithm to software ISP.
It is based on the corresponding algorithm in rkisp1.
The primary difference against hardware pipelines is that applying the
CCM is optional. Applying CCM causes a significant slowdown, time
needed to process a frame raises by 40-90% on tested platforms. If CCM
is really needed, it can be applied, if not, it's better to stick
without it. This can be configured by presence or omission of Ccm
algorithm in the tuning file.
CCM is changed only if the determined temperature changes by at least
100 K (an arbitrarily selected value), to avoid recomputing the matrices
and lookup tables all the time.
Since the CCM is float, rather than double, to use the same type as in
the rkisp1 pipeline, the type of color gains is changed from double to
float.
The outputs of the algorithm are not used yet, they will be enabled in
followup patches.
Signed-off-by: Milan Zamazal <mzamazal@redhat.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Milan Zamazal <mzamazal@redhat.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
The AWB algorithm has data to determine color temperature of the image.
Let's compute the temperature from it and store it into the context.
This piece of information is currently unused but it will be needed in a
followup patch introducing support for color correction matrix.
Let's store the white balance related information under `awb' subsection
of the active state, as the hardware pipelines do.
Signed-off-by: Milan Zamazal <mzamazal@redhat.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Milan Zamazal <mzamazal@redhat.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
The ExposureModeHelper::splitExposures() runs through the configured
stages to find the best gain/exposure time pair. It first raises the
exposure time until it reaches the limit of the current stage. Then it
raises the gain until that also reaches the limit of the current stage.
After that it continues with the next stage until a match is found.
Due to a slight mistake in the initial code, the second step doesn't
work as expected because the exposure time gets divided by the gain of
the current stage, effectively leading to a jump of the gain value from
the maximum gain of the last stage to the maximum gain of the current
stage instead of gradually increasing the gain value.
Depending on the tuning file this leads to very visible oscillations and
jumps in the brightness.
Fix by clamping the exposure time in the second step to the maximum
exposure time of the current stage.
While at it, add two comments for easier understanding.
Fixes: 34c9ab6282 ("ipa: libipa: Add ExposureModeHelper")
Signed-off-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
The minimum FrameDurationLimit also limits the min exposure time and
results in overly bright AE regulation. Remove the limit on the minimum
exposure time as the vertical blanking ensures the minimum frame
duration limit.
Fixes: f72c76eb6e ("rkisp1: Honor the FrameDurationLimits control")
Signed-off-by: Stefan Klug <stefan.klug@ideasonboard.com>
Reviewed-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Tested-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Reviewed-by: Jacopo Mondi <jacopo.mondi@ideasonboard.com>
The configured line duration of the sensor is used frequently throughout
the AGC implementation.
It's available in the IPA context through the rather long:
context.configuration.sensor.lineDuration
Take a copy of the lineDuration early in the call and replace the two
current usages of the reference with the shorter copy to manage line
length and ease readibility.
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
Signed-off-by: Paul Elder <paul.elder@ideasonboard.com>
Reviewed-by: Umang Jain <umang.jain@ideasonboard.com>
Reviewed-by: Daniel Scally <dan.scally@ideasonboard.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
