"""
ATD (1995) Colour Vision Model
==============================
Define the *ATD (1995)* colour vision model.
- :class:`colour.CAM_Specification_ATD95`
- :func:`colour.XYZ_to_ATD95`
Notes
-----
- According to *CIE TC1-34* definition of a colour appearance model, the
*ATD (1995)* model cannot be considered as a colour appearance model.
It was developed with different aims and is described as a model of
colour vision.
References
----------
- :cite:`Fairchild2013v` : Fairchild, M. D. (2013). ATD Model. In Color
Appearance Models (3rd ed., pp. 5852-5991). Wiley. ISBN:B00DAYO8E2
- :cite:`Guth1995a` : Guth, S. L. (1995). Further applications of the ATD
model for color vision. In E. Walowit (Ed.), Proc. SPIE 2414,
Device-Independent Color Imaging II (Vol. 2414, pp. 12-26).
doi:10.1117/12.206546
"""
from __future__ import annotations
import typing
from dataclasses import dataclass, field
from colour.algebra import sdiv, sdiv_mode, spow, vecmul
if typing.TYPE_CHECKING:
from colour.hints import Annotated, ArrayLike, Domain100, NDArrayFloat
from colour.utilities import (
MixinDataclassArithmetic,
array_namespace,
as_float,
from_range_degrees,
to_domain_100,
tsplit,
tstack,
xp_as_float_array,
)
__author__ = "Colour Developers"
__copyright__ = "Copyright 2013 Colour Developers"
__license__ = "BSD-3-Clause - https://opensource.org/licenses/BSD-3-Clause"
__maintainer__ = "Colour Developers"
__email__ = "colour-developers@colour-science.org"
__status__ = "Production"
__all__ = [
"CAM_ReferenceSpecification_ATD95",
"CAM_Specification_ATD95",
"XYZ_to_ATD95",
]
@dataclass
class CAM_ReferenceSpecification_ATD95(MixinDataclassArithmetic):
"""
Define the *ATD (1995)* colour vision model reference specification.
This specification contains field names consistent with the *Fairchild
(2013)* reference.
Parameters
----------
H
*Hue* angle :math:`H` in degrees.
C
Correlate of *saturation* :math:`C`. *Guth (1995)* incorrectly uses
the terms saturation and chroma interchangeably. However, :math:`C`
represents a measure of saturation rather than chroma since it is
calculated relative to the achromatic response for the stimulus
rather than that of a similarly illuminated white.
Br
Correlate of *brightness* :math:`Br`.
A_1
First stage :math:`A_1` response.
T_1
First stage :math:`T_1` response.
D_1
First stage :math:`D_1` response.
A_2
Second stage :math:`A_2` response.
T_2
Second stage :math:`A_2` response.
D_2
Second stage :math:`D_2` response.
References
----------
:cite:`Fairchild2013v`, :cite:`Guth1995a`
"""
H: float | NDArrayFloat | None = field(default_factory=lambda: None)
C: float | NDArrayFloat | None = field(default_factory=lambda: None)
Br: float | NDArrayFloat | None = field(default_factory=lambda: None)
A_1: float | NDArrayFloat | None = field(default_factory=lambda: None)
T_1: float | NDArrayFloat | None = field(default_factory=lambda: None)
D_1: float | NDArrayFloat | None = field(default_factory=lambda: None)
A_2: float | NDArrayFloat | None = field(default_factory=lambda: None)
T_2: float | NDArrayFloat | None = field(default_factory=lambda: None)
D_2: float | NDArrayFloat | None = field(default_factory=lambda: None)
[docs]
@dataclass
class CAM_Specification_ATD95(MixinDataclassArithmetic):
"""
Define the *ATD (1995)* colour vision model specification.
This specification provides a standardized interface for the *ATD (1995)*
model with field names consistent across all colour appearance models in
:mod:`colour.appearance`. While the field names differ from the original
*Fairchild (2013)* reference notation, they map directly to the model's
perceptual correlates.
Parameters
----------
h
*Hue* :math:`H = T_2 / D_2` per *Guth (1995)* Equation 14.24;
the raw opponent-channel ratio, not a hue angle in degrees. A
proper hue angle can be obtained from
:func:`numpy.arctan2`\\ ``(T_2, D_2)`` per *Fairchild (2013)*
p.243, which notes the raw ratio is equivocal (equal for
complementary hues, infinite or undefined in some cases).
C
Correlate of *saturation* :math:`C`. *Guth (1995)* incorrectly uses
the terms saturation and chroma interchangeably. However, :math:`C`
represents a measure of saturation rather than chroma since it is
measured relative to the achromatic response for the stimulus rather
than that of a similarly illuminated white.
Q
Correlate of *brightness* :math:`Br`.
A_1
First stage :math:`A_1` response.
T_1
First stage :math:`T_1` response.
D_1
First stage :math:`D_1` response.
A_2
Second stage :math:`A_2` response.
T_2
Second stage :math:`T_2` response.
D_2
Second stage :math:`D_2` response.
Notes
-----
- This specification is the one used in the current model
implementation.
References
----------
:cite:`Fairchild2013v`, :cite:`Guth1995a`
"""
h: float | NDArrayFloat | None = field(default_factory=lambda: None)
C: float | NDArrayFloat | None = field(default_factory=lambda: None)
Q: float | NDArrayFloat | None = field(default_factory=lambda: None)
A_1: float | NDArrayFloat | None = field(default_factory=lambda: None)
T_1: float | NDArrayFloat | None = field(default_factory=lambda: None)
D_1: float | NDArrayFloat | None = field(default_factory=lambda: None)
A_2: float | NDArrayFloat | None = field(default_factory=lambda: None)
T_2: float | NDArrayFloat | None = field(default_factory=lambda: None)
D_2: float | NDArrayFloat | None = field(default_factory=lambda: None)
[docs]
def XYZ_to_ATD95(
XYZ: Domain100,
XYZ_0: Domain100,
Y_0: ArrayLike,
k_1: ArrayLike,
k_2: ArrayLike,
sigma: ArrayLike = 300,
) -> Annotated[CAM_Specification_ATD95, 360]:
"""
Compute the *ATD (1995)* colour vision model correlates from the specified
*CIE XYZ* tristimulus values.
Parameters
----------
XYZ
*CIE XYZ* tristimulus values of test sample / stimulus.
XYZ_0
*CIE XYZ* tristimulus values of reference white.
Y_0
Absolute adapting field luminance in :math:`cd/m^2`.
k_1
Application specific weight :math:`k_1`.
k_2
Application specific weight :math:`k_2`.
sigma
Constant :math:`\\sigma` varied to predict different types of data.
Returns
-------
:class:`colour.CAM_Specification_ATD95`
*ATD (1995)* colour vision model specification.
Notes
-----
+---------------------+-----------------------+---------------+
| **Domain** | **Scale - Reference** | **Scale - 1** |
+=====================+=======================+===============+
| ``XYZ`` | 100 | 1 |
+---------------------+-----------------------+---------------+
| ``XYZ_0`` | 100 | 1 |
+---------------------+-----------------------+---------------+
+---------------------+-----------------------+---------------+
| **Range** | **Scale - Reference** | **Scale - 1** |
+=====================+=======================+===============+
| ``specification.h`` | 360 | 1 |
+---------------------+-----------------------+---------------+
- For unrelated colours, there is only self-adaptation and :math:`k_1`
is set to 1.0 while :math:`k_2` is set to 0.0. For related colours
such as typical colorimetric applications, :math:`k_1` is set to 0.0
and :math:`k_2` is set to a value between 15 and 50 *(Guth, 1995)*.
References
----------
:cite:`Fairchild2013v`, :cite:`Guth1995a`
Examples
--------
>>> import numpy as np
>>> XYZ = np.array([19.01, 20.00, 21.78])
>>> XYZ_0 = np.array([95.05, 100.00, 108.88])
>>> Y_0 = 318.31
>>> k_1 = 0.0
>>> k_2 = 50.0
>>> XYZ_to_ATD95(XYZ, XYZ_0, Y_0, k_1, k_2) # doctest: +ELLIPSIS
CAM_Specification_ATD95(h=np.float64(1.9089869...), \
C=np.float64(1.2064060...), Q=np.float64(0.1814003...), \
A_1=np.float64(0.1787931...), T_1=np.float64(0.0286942...), \
D_1=np.float64(0.0107584...), A_2=np.float64(0.0192182...), \
T_2=np.float64(0.0205377...), D_2=np.float64(0.0107584...))
"""
XYZ = to_domain_100(XYZ)
XYZ_0 = to_domain_100(XYZ_0)
xp = array_namespace(XYZ, XYZ_0, Y_0, k_1, k_2, sigma)
Y_0 = xp_as_float_array(Y_0, xp=xp, like=XYZ)
k_1 = xp_as_float_array(k_1, xp=xp, like=XYZ)
k_2 = xp_as_float_array(k_2, xp=xp, like=XYZ)
sigma = xp_as_float_array(sigma, xp=xp, like=XYZ)
# Converting luminance in :math:`cd/m^2` to retinal illuminance in trolands
# for the stimulus and the reference white.
XYZ = 18 * spow(Y_0[..., None] * XYZ / 100, 0.8)
XYZ_0 = 18 * spow(Y_0[..., None] * XYZ_0 / 100, 0.8)
# Computing the adaptation stimulus :math:`XYZ_a` then deriving the
# post-adaptation cone signals via the *ATD95* :math:`XYZ \\rightarrow LMS`
# transform applied to both the stimulus and the adaptation stimulus.
XYZ_a = k_1[..., None] * XYZ + k_2[..., None] * XYZ_0
LMS_scales = xp_as_float_array([0.66, 1.0, 0.43], xp=xp, like=XYZ)
LMS_offsets = xp_as_float_array([0.024, 0.036, 0.31], xp=xp, like=XYZ)
LMS_matrix = [
[0.2435, 0.8524, -0.0516],
[-0.3954, 1.1642, 0.0837],
[0.0000, 0.0400, 0.6225],
]
LMS = spow(vecmul(LMS_matrix, XYZ) * LMS_scales, 0.7) + LMS_offsets
LMS_a = spow(vecmul(LMS_matrix, XYZ_a) * LMS_scales, 0.7) + LMS_offsets
LMS_g = LMS * (sigma[..., None] / (sigma[..., None] + LMS_a))
# Computing opponent colour dimensions: 6 linear combinations of the
# post-adaptation cone signals, each passed through the saturating final
# response :math:`v / (200 + |v|)`.
L_g, M_g, S_g = tsplit(LMS_g)
A_1i = 3.57 * L_g + 2.64 * M_g
T_1i = 7.18 * L_g - 6.21 * M_g
D_1i = -0.7 * L_g + 0.085 * M_g + S_g
A_2i = 0.09 * A_1i
T_2i = 0.43 * T_1i + 0.76 * D_1i
D_2i = D_1i
stage = tstack([A_1i, T_1i, D_1i, A_2i, T_2i, D_2i])
stage_final = stage / (200 + xp.abs(stage))
A_1, T_1, D_1, A_2, T_2, D_2 = tsplit(stage_final)
# Computing the correlate of *brightness* :math:`Br`.
Br = spow(A_1**2 + T_1**2 + D_1**2, 0.5)
# Computing the correlate of *saturation* :math:`C` and the *hue*
# :math:`H`. Note that the reference does not take the modulus of the
# :math:`H`, thus :math:`H` can exceed 360 degrees.
with sdiv_mode():
C = sdiv(spow(T_2**2 + D_2**2, 0.5), A_2)
H = sdiv(T_2, D_2)
return CAM_Specification_ATD95(
h=as_float(from_range_degrees(H)),
C=as_float(C),
Q=as_float(Br),
A_1=as_float(A_1),
T_1=as_float(T_1),
D_1=as_float(D_1),
A_2=as_float(A_2),
T_2=as_float(T_2),
D_2=as_float(D_2),
)