#!/usr/bin/env python
# -*- coding: utf-8 -*-
"""
Clipped Highlights Recovery
===========================
Defines the clipped highlights recovery objects:
- :func:`highlights_recovery_blend`
See Also
--------
`Colour - HDRI - Examples: Merge from Raw Files IPython Notebook
<https://github.com/colour-science/colour-hdri/\
blob/master/colour_hdri/examples/examples_merge_from_raw_files.ipynb>`_
"""
from __future__ import division, unicode_literals
import numpy as np
from colour import dot_vector
__author__ = 'Colour Developers'
__copyright__ = 'Copyright (C) 2015-2016 - Colour Developers'
__license__ = 'New BSD License - http://opensource.org/licenses/BSD-3-Clause'
__maintainer__ = 'Colour Developers'
__email__ = 'colour-science@googlegroups.com'
__status__ = 'Production'
__all__ = ['highlights_recovery_blend']
[docs]def highlights_recovery_blend(RGB, multipliers, threshold=0.99):
"""
Performs highlights recovery using Coffin (1997) method from *dcraw*.
Parameters
----------
RGB : array_like
*RGB* colourspace array.
multipliers : array_like
Normalised camera white level or white balance multipliers.
threshold : numeric, optional
Threshold for the highlights selection.
Returns
-------
ndarray
Highlights recovered *RGB* colourspace array.
References
----------
.. [1] Coffin, D. (2015). dcraw. Retrieved from
https://www.cybercom.net/~dcoffin/dcraw/
"""
M = np.array([[1.0000000, 1.0000000, 1.0000000],
[1.7320508, -1.7320508, 0.0000000],
[-1.0000000, -1.0000000, 2.0000000]])
clipping_level = np.min(multipliers) * threshold
Lab = dot_vector(M, RGB)
Lab_c = dot_vector(M, np.minimum(RGB, clipping_level))
s = np.sum((Lab * Lab)[..., 1:3], axis=2)
s_c = np.sum((Lab_c * Lab_c)[..., 1:3], axis=2)
ratio = np.sqrt(s_c / s)
ratio[np.logical_or(np.isnan(ratio), np.isinf(ratio))] = 1
Lab[:, :, 1:3] *= np.rollaxis(ratio[np.newaxis], 0, 3)
RGB_o = dot_vector(np.linalg.inv(M), Lab)
return RGB_o