[PATCH v3 2/3] ipa: libipa: Add MatrixInterpolator class
Stefan Klug
stefan.klug at ideasonboard.com
Fri May 17 09:56:58 CEST 2024
Hi Paul,
On Fri, May 17, 2024 at 12:50:50PM +0900, Paul Elder wrote:
> On Wed, May 15, 2024 at 11:20:56AM +0200, Stefan Klug wrote:
> > On Tue, May 07, 2024 at 03:10:15PM +0900, Paul Elder wrote:
> > > Add a class to encapsulate the functionality of fetching a matrix based
> > > on an integer key, and interpolating if there is no exact match. This is
> > > expected to be used by both color correction matrices / crosstalk
> > > correction as well as lens shading correction.
> > >
> > > Signed-off-by: Paul Elder <paul.elder at ideasonboard.com>
> > >
> > > ---
> > > Changes in v3:
> > > - add a constructor that takes a map of unsigned int -> matrix
> > > - s/unit/identity
> > > - clear matrices on reset and when reading from yaml
> > > - add assert at get()
> > >
> > > Changes in v2:
> > > - initialize to identity matrix
> > > - add a function to reset to identity matrix
> > > - other minor fixes
> > > ---
> > > src/ipa/libipa/matrix_interpolator.cpp | 54 ++++++++++++
> > > src/ipa/libipa/matrix_interpolator.h | 116 +++++++++++++++++++++++++
> > > src/ipa/libipa/meson.build | 2 +
> > > 3 files changed, 172 insertions(+)
> > > create mode 100644 src/ipa/libipa/matrix_interpolator.cpp
> > > create mode 100644 src/ipa/libipa/matrix_interpolator.h
> > >
> > > diff --git a/src/ipa/libipa/matrix_interpolator.cpp b/src/ipa/libipa/matrix_interpolator.cpp
> > > new file mode 100644
> > > index 00000000..fc1837e7
> > > --- /dev/null
> > > +++ b/src/ipa/libipa/matrix_interpolator.cpp
> > > @@ -0,0 +1,54 @@
> > > +/* SPDX-License-Identifier: BSD-2-Clause */
> > > +/*
> > > + * Copyright (C) 2019, Raspberry Pi Ltd
> > > + * Copyright (C) 2024, Paul Elder <paul.elder at ideasonboard.com>
> > > + *
> > > + * matrix_interpolator.cpp - Helper class for interpolating maps of matrices
> > > + */
> > > +#include "matrix_interpolator.h"
> > > +
> > > +#include <algorithm>
> > > +#include <string>
> > > +
> > > +#include <libcamera/base/log.h>
> > > +
> > > +#include "libcamera/internal/yaml_parser.h"
> > > +
> > > +#include "matrix.h"
> > > +
> > > +/**
> > > + * \file ccm.h
> > > + * \brief Helper class for interpolating maps of matrices
> > > + */
> > > +
> > > +namespace libcamera {
> > > +
> > > +LOG_DEFINE_CATEGORY(MatrixInterpolator)
> > > +
> > > +namespace ipa {
> > > +
> > > +/**
> > > + * \class MatrixInterpolator
> > > + * \brief Class for storing, retrieving, and interpolating matrices
> > > + */
> > > +
> > > +/**
> > > + * \fn int MatrixInterpolator<T, R, C>::readYaml(const libcamera::YamlObject &yaml)
> > > + * \brief Initialize an MatrixInterpolator instance from yaml
> > > + * \tparam T Type of data stored in the matrices
> > > + * \tparam R Number of rows of the matrices
> > > + * \tparam C Number of columns of the matrices
> > > + * \param[in] yaml The yaml object that contains the map of unsigned integers to matrices
> > > + * \return Zero on success, negative error code otherwise
> > > + */
> > > +
> > > +/**
> > > + * \fn Matrix<T, R, C> MatrixInterpolator<T, R, C>::get(unsigned int key)
> > > + * \brief Retrieve a matrix from the list of matrices, interpolating if necessary
> > > + * \param[in] key The unsigned integer key of the matrix to retrieve
> > > + * \return The matrix corresponding to the color temperature
> > > + */
> > > +
> > > +} /* namespace ipa */
> > > +
> > > +} /* namespace libcamera */
> > > diff --git a/src/ipa/libipa/matrix_interpolator.h b/src/ipa/libipa/matrix_interpolator.h
> > > new file mode 100644
> > > index 00000000..2c6263cb
> > > --- /dev/null
> > > +++ b/src/ipa/libipa/matrix_interpolator.h
> > > @@ -0,0 +1,116 @@
> > > +/* SPDX-License-Identifier: BSD-2-Clause */
> > > +/*
> > > + * Copyright (C) 2019, Raspberry Pi Ltd
> > > + * Copyright (C) 2024, Paul Elder <paul.elder at ideasonboard.com>
> > > + *
> > > + * matrix_interpolator.h - Helper class for interpolating maps of matrices
> > > + */
> > > +
> > > +#pragma once
> > > +
> > > +#include <algorithm>
> > > +#include <map>
> > > +#include <string>
> > > +#include <tuple>
> > > +
> > > +#include <libcamera/base/log.h>
> > > +
> > > +#include "libcamera/internal/yaml_parser.h"
> > > +
> > > +#include "matrix.h"
> > > +
> > > +namespace libcamera {
> > > +
> > > +LOG_DECLARE_CATEGORY(MatrixInterpolator)
> > > +
> > > +namespace ipa {
> > > +
> > > +template<typename T, unsigned int R, unsigned int C,
> > > + std::enable_if_t<std::is_arithmetic_v<T>> * = nullptr>
> > > +class MatrixInterpolator
> > > +{
> > > +public:
> > > + MatrixInterpolator()
> > > + {
> > > + reset();
> > > + };
> > > +
> > > + MatrixInterpolator(const std::map<unsigned int, Matrix<T, R, C>> &matrices)
> > > + {
> > > + for (const auto &pair : matrices)
> > > + matrices_[pair.first] = pair.second;
> > > + };
> > > +
> > > + ~MatrixInterpolator(){};
> > > +
> > > + void reset()
> > > + {
> > > + Matrix<T, R, C> identity;
> > > + matrices_.clear();
> > > + matrices_[0] = identity;
> > > + }
> > > +
> > > + int readYaml(const libcamera::YamlObject &yaml)
> > > + {
> > > + int ret;
> > > +
> > > + matrices_.clear();
> > > +
> > > + if (!yaml.isDictionary()) {
> > > + LOG(MatrixInterpolator, Error) << "yaml object must be a dictionary";
> > > + return -EINVAL;
> > > + }
> > > +
> >
> > I have the feeling that this might not be flexible enough. At the moment
> > this expects a yaml representation of
> >
> > 2000: [ matrix values... ]
> > 3000: [ ... ]
> >
> > We might need to add more infos for a given color temperature. In my
> > prototypes I used a list like this:
> >
> > - ct: 2000
> > ccm: [matrix values]
> > - ct: 3000
> > ccm: [ matrix values ]
>
> Yes, I think that will be necessary, also to match the layout with lsc
> so that we can use this for lsc in the future.
>
> >
> >
> > This adds the flexibility to e.g. add the offset vector.
> > My diff is:
> > - int readYaml(const libcamera::YamlObject &yaml)
> > + int readYaml(const libcamera::YamlObject &yaml, const std::string& key_property, const std::string& value_property)
>
> Ah, good idea. I was trying to figure out how to support different size
> matrices in the same map...
>
> > {
> > int ret;
> >
> > matrices_.clear();
> >
> > - if (!yaml.isDictionary()) {
> > - LOG(MatrixInterpolator, Error) << "yaml object must be a dictionary";
> > + if (!yaml.isList()) {
> > + LOG(MatrixInterpolator, Error) << "yaml object must be a list";
> > return -EINVAL;
> > }
> >
> > - for (const auto &[ctStr, value] : yaml.asDict()) {
> > - unsigned int ct = std::stoul(ctStr);
> > + for (const auto &value : yaml.asList()) {
> > + unsigned int ct = std::stoul(value[key_property].get<std::string>(""));
> > Matrix<T, R, C> matrix;
> > - if ((ret = matrix.readYaml(value)) < 0) {
> > + if ((ret = matrix.readYaml(value[value_property])) < 0) {
> > LOG(MatrixInterpolator, Error) << "Failed to read matrix";
> > return ret;
> > }
> >
> >
> >
> > I'm not yet sure where we will end up. This largely depends on how the
> > tuning file format evolves. What are your thoughts about that?
> >
> > > + for (const auto &[ctStr, value] : yaml.asDict()) {
> > > + unsigned int ct = std::stoul(ctStr);
> > > + Matrix<T, R, C> matrix;
> > > + if ((ret = matrix.readYaml(value)) < 0) {
> > > + LOG(MatrixInterpolator, Error) << "Failed to read matrix";
> > > + return ret;
> > > + }
> > > +
> > > + matrices_[ct] = matrix;
> > > +
> > > + LOG(MatrixInterpolator, Debug)
> > > + << "Read matrix for key " << ct << ": "
> > > + << matrices_[ct].toString();
> > > + }
> > > +
> > > + if (matrices_.size() < 1) {
> > > + LOG(MatrixInterpolator, Error) << "Need at least one matrix";
> > > + return -EINVAL;
> > > + }
> > > +
> > > + return 0;
> > > + }
> > > +
> > > + Matrix<T, R, C> get(unsigned int ct)
> > > + {
> > > + ASSERT(matrices_.size() > 0);
> >
> > I'm a bit concerned here. This check only runs in debug mode. But in
> > case the readYaml gets an invalid object, it leaves the interpolator
> > with matrices_.size() == 0 and we would crash here. I think we should at
> > least catch that in release mode and maybe return a identity matrix (and
> > then there is need to inject the identity in the constructor anymore)
>
> tbh I think it should be fine because it'll be caught in readYaml and we
> return -EINVAL. Is that not sufficient?
Yes, that's right. The error case should be handled in the layers above.
And after the -EINVAL this function *should* not be called. So I think
it's just personal preference (personally I would throw here, which is
equally drastic as getting a possible segfault but more deterministic :-).
So it's ok for me.
Cheers,
Stefan
>
>
> Thanks,
>
> Paul
>
> >
> > > +
> > > + if (matrices_.size() == 1 ||
> > > + ct <= matrices_.begin()->first)
> > > + return matrices_.begin()->second;
> > > +
> > > + if (ct >= matrices_.rbegin()->first)
> > > + return matrices_.rbegin()->second;
> > > +
> > > + if (matrices_.count(ct))
> > > + return matrices_[ct];
> > > +
> > > + /* The above four guarantee that this will succeed */
> > > + auto iter = matrices_.upper_bound(ct);
> > > + unsigned int ctUpper = iter->first;
> > > + unsigned int ctLower = (--iter)->first;
> > > +
> > > + double lambda = (ct - ctLower) / static_cast<double>(ctUpper - ctLower);
> > > + return lambda * matrices_[ctUpper] + (1.0 - lambda) * matrices_[ctLower];
> > > + }
> > > +
> > > +private:
> > > + std::map<unsigned int, Matrix<T, R, C>> matrices_;
> > > +};
> > > +
> > > +} /* namespace ipa */
> > > +
> > > +} /* namespace libcamera */
> > > diff --git a/src/ipa/libipa/meson.build b/src/ipa/libipa/meson.build
> > > index 1e34355f..305acf64 100644
> > > --- a/src/ipa/libipa/meson.build
> > > +++ b/src/ipa/libipa/meson.build
> > > @@ -8,6 +8,7 @@ libipa_headers = files([
> > > 'fc_queue.h',
> > > 'histogram.h',
> > > 'matrix.h',
> > > + 'matrix_interpolator.h',
> > > 'module.h',
> > > 'pwl.h',
> > > ])
> > > @@ -20,6 +21,7 @@ libipa_sources = files([
> > > 'fc_queue.cpp',
> > > 'histogram.cpp',
> > > 'matrix.cpp',
> > > + 'matrix_interpolator.cpp',
> > > 'module.cpp',
> > > 'pwl.cpp'
> > > ])
> > > --
> > > 2.39.2
> > >
More information about the libcamera-devel
mailing list