[libcamera-devel] [PATCH 3/3] libcamera: qcam: Improve colour information in DNG files

[Laurent Pinchart]

Hi David,

Thank you for the patch.

On Sat, Jul 04, 2020 at 10:59:14AM +0100, David Plowman wrote:
> This patch improves the colour information recorded in DNG files using
> the ColourCorrectionMatrix metadata for the image. Note that we are
> not supplying a full calibration using two illuminants, nonetheless
> the single matrix here appears to be respected by a number of tools.
> 
> Signed-off-by: David Plowman <david.plowman at raspberrypi.com>
> ---
>  src/qcam/dng_writer.cpp | 93 +++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 93 insertions(+)
> 
> diff --git a/src/qcam/dng_writer.cpp b/src/qcam/dng_writer.cpp
> index 61505d3..222df9f 100644
> --- a/src/qcam/dng_writer.cpp
> +++ b/src/qcam/dng_writer.cpp
> @@ -34,6 +34,63 @@ struct FormatInfo {
>  			      unsigned int stride);
>  };
>  
> +struct Matrix {

I'd name this Matrix3d as it's a 3x3 matrix.

> +	Matrix(float m0, float m1, float m2,
> +	       float m3, float m4, float m5,
> +	       float m6, float m7, float m8)
> +	{
> +		m[0] = m0, m[1] = m1, m[2] = m2;
> +		m[3] = m3, m[4] = m4, m[5] = m5;
> +		m[6] = m6, m[7] = m7, m[8] = m8;
> +	}

Please add blank lines between functions.

> +	Matrix(float diag0, float diag1, float diag2)
> +		: Matrix(diag0, 0, 0, 0, diag1, 0, 0, 0, diag2) {}

And we tend to move the brackets to lines of their own after the member
initializer list.

I wonder if this should be a static function though,

	static Matrix diag(float diag0, float diag1, float diag2)
	{
		return { diag0, 0, 0, 0, diag1, 0, 0, 0, diag2 };
	}

as using a custom constructor is a bit confusing. When I was reading the
code below:

	Matrix wbGain(1, 1, 1);

I wondered how a 3x3 matrix would be constructed from 3 values only.

	Matrix wbGain = Matrix::diag(1, 1, 1);

would be more explicit. Another useful function would be

	Matrix identity()
	{
		return { 1, 0, 0, 0, 1, 0, 0, 0, 1 };
	}

> +	Matrix() {}

We usually put the default constructor first.

> +	float m[9];

We usually put member data after member functions.

> +	Matrix transpose() const
> +	{
> +		return Matrix(m[0], m[3], m[6], m[1], m[4], m[7], m[2], m[5], m[8]);

You can also write

		return { m[0], m[3], m[6], m[1], m[4], m[7], m[2], m[5], m[8] };

which will cause the compiler to warn in case of loss of precision due
to implicit casts. Same for cofactors().

> +	}
> +	Matrix cofactors() const
> +	{
> +		return Matrix(m[4] * m[8] - m[5] * m[7],
> +			      -(m[3] * m[8] - m[5] * m[6]),
> +			      m[3] * m[7] - m[4] * m[6],
> +			      -(m[1] * m[8] - m[2] * m[7]),
> +			      m[0] * m[8] - m[2] * m[6],
> +			      -(m[0] * m[7] - m[1] * m[6]),
> +			      m[1] * m[5] - m[2] * m[4],
> +			      -(m[0] * m[5] - m[2] * m[3]),
> +			      m[0] * m[4] - m[1] * m[3]);
> +	}
> +	Matrix adjugate() const { return cofactors().transpose(); }
> +	float determinant() const
> +	{
> +		return (m[0] * (m[4] * m[8] - m[5] * m[7]) -
> +			m[1] * (m[3] * m[8] - m[5] * m[6]) +
> +			m[2] * (m[3] * m[7] - m[4] * m[6]));

The outer parentheses are not needed.

> +	}
> +	Matrix inverse() const { return adjugate() * (1.0 / determinant()); }

Should we protect against determinant being 0 ?

> +	Matrix operator*(Matrix const &other) const
> +	{
> +		Matrix result;
> +		for (int i = 0; i < 3; i++)

unsigned int i

Even though not strictly necessary, we use curly braces when the inner
statement is a compound statement.

> +			for (int j = 0; j < 3; j++)

unsigned int j

> +				result.m[i * 3 + j] =
> +					m[i * 3 + 0] * other.m[0 + j] +
> +					m[i * 3 + 1] * other.m[3 + j] +
> +					m[i * 3 + 2] * other.m[6 + j];
> +		return result;
> +	}
> +	Matrix operator*(float const &f) const

I think "float f" would be fine, is there a need for a reference ?

> +	{
> +		Matrix result;
> +		for (int i = 0; i < 9; i++)

unsigned int i

> +			result.m[i] = m[i] * f;
> +		return result;
> +	}
> +};
> +
>  void packScanlineSBGGR10P(void *output, const void *input, unsigned int width)
>  {
>  	const uint8_t *in = static_cast<const uint8_t *>(input);
> @@ -315,6 +372,42 @@ int DNGWriter::write(const char *filename, const Camera *camera,
>  	TIFFSetField(tif, TIFFTAG_PLANARCONFIG, PLANARCONFIG_CONTIG);
>  	TIFFSetField(tif, TIFFTAG_SAMPLEFORMAT, SAMPLEFORMAT_UINT);
>  
> +	/*
> +	 * Fill in some reasonable colour information in the DNG. We supply
> +	 * the "neutral" colour values which determine the white balance, and the
> +	 * "ColorMatrix1" which converts XYZ to (un-white-balanced) camera RGB.
> +	 * Note that this is not a "proper" colour calibration for the DNG,
> +	 * nonetheless, many tools should be able to render the colours better.
> +	 */
> +	float neutral[3] = { 1, 1, 1 };
> +	Matrix wbGain(1, 1, 1);
> +	/* From http://www.brucelindbloom.com/index.html?Eqn_RGB_XYZ_Matrix.html */

https would be better, but it seems that the server doesn't support it
:-(

> +	Matrix rgb2xyz(0.4124564, 0.3575761, 0.1804375,
> +		       0.2126729, 0.7151522, 0.0721750,
> +		       0.0193339, 0.1191920, 0.9503041);

This should be const.

> +	Matrix ccm(1, 1, 1);
> +	const double eps = 1e-2;

Out of curiosity, why 1e-2 ?

> +
> +	if (metadata.contains(controls::ColourGains)) {
> +		Span<const float> colour_gains = metadata.get(controls::ColourGains);

s/colour_gains/colourGains/ or just s/colour_gains/gains/

> +		if (colour_gains[0] > eps && colour_gains[1] > eps) {
> +			wbGain = Matrix(colour_gains[0], 1, colour_gains[1]);
> +			neutral[0] = 1.0 / colour_gains[0]; /* red */
> +			neutral[2] = 1.0 / colour_gains[1]; /* blue */
> +		}
> +	}
> +	if (metadata.contains(controls::ColourCorrectionMatrix)) {
> +		Span<const float> m = metadata.get(controls::ColourCorrectionMatrix);
> +		Matrix tmp = Matrix(m[0], m[1], m[2], m[3], m[4], m[5], m[6], m[7], m[8]);

Would a

	Matrix(Span<const float> m)

constructor make sense to simplify this line ? Up to you.

We really try not to use "tmp" as a variable name. Possible alternatives
are "matrix", or just "m" if you rename the existing "m" to "value".

> +		if (tmp.determinant() > eps)
> +			ccm = tmp;
> +	}

Please add a blank line here.

> +	/* This is guaranteed to be invertible because all the bits in it are. */

I'd expand this a little:

	/*
	 * This is guaranteed to be invertible because all the bits in it are
	 * (rgb2xyz is hardcoded and invertible, the ccm determinant is checked
	 * manually above, and wbGain is a diagonal matrix with diagonal
	 * elements checked to be non-zero).
	 */

> +	Matrix colorMatrix1 = (rgb2xyz * ccm * wbGain).inverse();
> +
> +	TIFFSetField(tif, TIFFTAG_COLORMATRIX1, 9, colorMatrix1.m);
> +	TIFFSetField(tif, TIFFTAG_ASSHOTNEUTRAL, 3, neutral);
> +
>  	/*
>  	 * Reserve space for the SubIFD and ExifIFD tags, pointing to the IFD
>  	 * for the raw image and EXIF data respectively. The real offsets will

-- 
Regards,

Laurent Pinchart