def dot23(A,B, keepdims = False):
"""
Dot product of a 2-d ndarray with a (N x K x L) 3-d ndarray
using einsum().
Args:
:A:
| ndarray (.shape = (M,N))
:B:
| ndarray (.shape = (N,K,L))
Returns:
:returns:
| ndarray (.shape = (M,K,L))
"""
if (len(A.shape)==2) & (len(B.shape)==3):
dotAB = np.einsum('ij,jkl->ikl',A,B)
if (len(B.shape)==3) & (keepdims == True):
dotAB = np.expand_dims(dotAB,axis=1)
elif (len(A.shape)==2) & (len(B.shape)==2):
dotAB = np.einsum('ij,jk->ik',A,B)
if (len(B.shape)==2) & (keepdims == True):
dotAB = np.expand_dims(dotAB,axis=1)
return dotAB
def _dot_M_xyz(M,xyz):
"""
Perform matrix multiplication between M and xyz (M*xyz) using einsum.
Args:
:xyz:
| 2D or 3D ndarray
:M:
| 2D or 3D ndarray
| if 2D: use same matrix M for each xyz,
| if 3D: use M[i] for xyz[i,:] (2D xyz) or xyz[:,i,:] (3D xyz)
Returns:
:M*xyz:
| ndarray with same shape as xyz containing dot product of M and xyz.
"""
# convert xyz to ...:
if np.ndim(M)==2:
if len(xyz.shape) == 3:
return np.einsum('ij,klj->kli', M, xyz)
else:
return np.einsum('ij,lj->li', M, xyz)
else:
if len(xyz.shape) == 3: # second dim of x must match dim of 1st of M and 1st dim of xyzw
return np.concatenate([np.einsum('ij,klj->kli', M[i], xyz[:,i:i+1,:]) for i in range(M.shape[0])],axis=1)
else: # first dim of xyz must match dim of 1st of M and 1st dim of xyzw
return np.concatenate([np.einsum('ij,lj->li', M[i], xyz[i:i+1,:]) for i in range(M.shape[0])],axis=0)
def lms_to_xyz(lms, cieobs=_CIEOBS, M=None, **kwargs):
"""
Convert LMS cone fundamental responses to XYZ tristimulus values.
Args:
:lms:
| ndarray with LMS cone fundamental responses
:cieobs:
| _CIEOBS or str, optional
:M:
| None, optional
| Conversion matrix for xyz to lms.
| If None: use the one defined by :cieobs:
Returns:
:xyz:
| ndarray with tristimulus values
"""
lms = np2d(lms)
if M is None:
M = _CMF[cieobs]['M']
# convert from lms to xyz:
if len(lms.shape) == 3:
xyz = np.einsum('ij,klj->kli', np.linalg.inv(M), lms)
else:
xyz = np.einsum('ij,lj->li', np.linalg.inv(M), lms)
return xyz
def xyz_to_lms(xyz, cieobs=_CIEOBS, M=None, **kwargs):
"""
Convert XYZ tristimulus values to LMS cone fundamental responses.
Args:
:xyz:
| ndarray with tristimulus values
:cieobs:
| _CIEOBS or str, optional
:M:
| None, optional
| Conversion matrix for xyz to lms.
| If None: use the one defined by :cieobs:
Returns:
:lms:
| ndarray with LMS cone fundamental responses
"""
xyz = np2d(xyz)
if M is None:
M = _CMF[cieobs]['M']
# convert xyz to lms:
if len(xyz.shape) == 3:
lms = np.einsum('ij,klj->kli', M, xyz)
else:
lms = np.einsum('ij,lj->li', M, xyz)
return lms
def Vrb_mb_to_xyz(Vrb,
cieobs=_CIEOBS,
scaling=[1, 1],
M=None,
Minverted=False,
**kwargs):
"""
Convert V,r,b (Macleod-Boynton) color coordinates to XYZ tristimulus values.
| Macleod Boynton: V = R+G, r = R/V, b = B/V
| Note that R,G,B ~ L,M,S
Args:
:Vrb:
| ndarray with V,r,b (Macleod-Boynton) color coordinates
:cieobs:
| luxpy._CIEOBS, optional
| CMF set to use when getting the default M, which is
| the xyz to lms conversion matrix.
:scaling:
| list of scaling factors for r and b dimensions.
:M:
| None, optional
| Conversion matrix for going from XYZ to RGB (LMS)
| If None, :cieobs: determines the M (function does inversion)
:Minverted:
| False, optional
| Bool that determines whether M should be inverted.
Returns:
:xyz:
| ndarray with tristimulus values
Reference:
1. `MacLeod DI, and Boynton RM (1979).
Chromaticity diagram showing cone excitation by stimuli of equal luminance.
J. Opt. Soc. Am. 69, 1183–1186.
<https://www.osapublishing.org/josa/abstract.cfm?uri=josa-69-8-1183>`_
"""
Vrb = np2d(Vrb)
RGB = np.empty(Vrb.shape)
RGB[..., 0] = Vrb[..., 1] * Vrb[..., 0] / scaling[0]
RGB[..., 2] = Vrb[..., 2] * Vrb[..., 0] / scaling[1]
RGB[..., 1] = Vrb[..., 0] - RGB[..., 0]
if M is None:
M = _CMF[cieobs]['M']
if Minverted == False:
M = np.linalg.inv(M)
if len(RGB.shape) == 3:
return np.einsum('ij,klj->kli', M, RGB)
else:
return np.einsum('ij,lj->li', M, RGB)
def ndinterp1(X, Y, Xnew):
"""
Perform nd-dimensional linear interpolation using Delaunay triangulation.
Args:
:X:
| ndarray with n-dimensional coordinates (last axis represents dimension).
:Y:
| ndarray with values at coordinates in X.
:Xnew:
| ndarray of new coordinates (last axis represents dimension).
| When outside of the convex hull of X, then a best estimate is
| given based on the closest vertices.
Returns:
:Ynew:
| ndarray with new values at coordinates in Xnew.
"""
#get dimensions:
n = Xnew.shape[-1]
# create an object with triangulation
tri = sp.spatial.Delaunay(X)
# find simplexes that contain interpolated points
s = tri.find_simplex(Xnew)
# get the vertices for each simplex
v = tri.vertices[s]
# get transform matrices for each simplex (see explanation bellow)
m = tri.transform[s]
# for each interpolated point p, mutliply the transform matrix by
# vector p-r, where r=m[:,n,:] is one of the simplex vertices to which
# the matrix m is related to (again, see below)
b = np.einsum('ijk,ik->ij', m[:,:n,:n], Xnew-m[:,n,:])
# get the weights for the vertices; `b` contains an n-dimensional vector
# with weights for all but the last vertices of the simplex
# (note that for n-D grid, each simplex consists of n+1 vertices);
# the remaining weight for the last vertex can be copmuted from
# the condition that sum of weights must be equal to 1
w = np.c_[b, 1-b.sum(axis=1)]
# normalize weigths:
w = w/w.sum(axis=1, keepdims=True)
# interpolate:
if Y[v].ndim == 3:
Ynew = np.einsum('ijk,ij->ik', Y[v], w)
else:
Ynew = np.einsum('ij,ij->i', Y[v], w)
return Ynew
def xyz_to_srgb(xyz, gamma=2.4, **kwargs):
"""
Calculates IEC:61966 sRGB values from xyz.
Args:
:xyz:
| ndarray with relative tristimulus values.
:gamma:
| 2.4, optional
| compression in sRGB
Returns:
:rgb:
| ndarray with R,G,B values (uint8).
"""
xyz = np2d(xyz)
# define 3x3 matrix
M = np.array([[3.2404542, -1.5371385, -0.4985314],
[-0.9692660, 1.8760108, 0.0415560],
[0.0556434, -0.2040259, 1.0572252]])
if len(xyz.shape) == 3:
srgb = np.einsum('ij,klj->kli', M, xyz / 100)
else:
srgb = np.einsum('ij,lj->li', M, xyz / 100)
# perform clipping:
srgb[np.where(srgb > 1)] = 1
srgb[np.where(srgb < 0)] = 0
# test for the dark colours in the non-linear part of the function:
dark = np.where(srgb <= 0.0031308)
# apply gamma function:
g = 1 / gamma
# and scale to range 0-255:
rgb = srgb.copy()
rgb = (1.055 * rgb**g - 0.055) * 255
# non-linear bit for dark colours
rgb[dark] = (srgb[dark].copy() * 12.92) * 255
# clip to range:
rgb[rgb > 255] = 255
rgb[rgb < 0] = 0
return rgb
def xyz_to_Vrb_mb(xyz, cieobs=_CIEOBS, scaling=[1, 1], M=None, **kwargs):
"""
Convert XYZ tristimulus values to V,r,b (Macleod-Boynton) color coordinates.
| Macleod Boynton: V = R+G, r = R/V, b = B/V
| Note that R,G,B ~ L,M,S
Args:
:xyz:
| ndarray with tristimulus values
:cieobs:
| luxpy._CIEOBS, optional
| CMF set to use when getting the default M, which is
the xyz to lms conversion matrix.
:scaling:
| list of scaling factors for r and b dimensions.
:M:
| None, optional
| Conversion matrix for going from XYZ to RGB (LMS)
| If None, :cieobs: determines the M (function does inversion)
Returns:
:Vrb:
| ndarray with V,r,b (Macleod-Boynton) color coordinates
Reference:
1. `MacLeod DI, and Boynton RM (1979).
Chromaticity diagram showing cone excitation by stimuli of equal luminance.
J. Opt. Soc. Am. 69, 1183–1186.
<https://www.osapublishing.org/josa/abstract.cfm?uri=josa-69-8-1183>`_
"""
xyz = np2d(xyz)
if M is None:
M = _CMF[cieobs]['M']
if len(xyz.shape) == 3:
RGB = np.einsum('ij,klj->kli', M, xyz)
else:
RGB = np.einsum('ij,lj->li', M, xyz)
Vrb = np.empty(xyz.shape)
Vrb[..., 0] = RGB[..., 0] + RGB[..., 1]
Vrb[..., 1] = RGB[..., 0] / Vrb[..., 0] * scaling[0]
Vrb[..., 2] = RGB[..., 2] / Vrb[..., 0] * scaling[1]
return Vrb
def srgb_to_xyz(rgb, gamma=2.4, **kwargs):
"""
Calculates xyz from IEC:61966 sRGB values.
Args:
:rgb:
| ndarray with srgb values (uint8).
:gamma:
| 2.4, optional
| compression in sRGB
Returns:
:xyz:
| ndarray with relative tristimulus values.
"""
rgb = np2d(rgb)
# define 3x3 matrix
M = np.array([[0.4124564, 0.3575761, 0.1804375],
[0.2126729, 0.7151522, 0.0721750],
[0.0193339, 0.1191920, 0.9503041]])
# scale device coordinates:
sRGB = rgb / 255
# test for non-linear part of conversion
nonlin = np.where((rgb / 255) < 0.0031308) #0.03928)
# apply gamma function to convert to sRGB
srgb = sRGB.copy()
srgb = ((srgb + 0.055) / 1.055)**gamma
srgb[nonlin] = sRGB[nonlin] / 12.92
if len(srgb.shape) == 3:
xyz = np.einsum('ij,klj->kli', M, srgb) * 100
else:
xyz = np.einsum('ij,lj->li', M, srgb) * 100
return xyz
def _cij_to_gij(xyz,C):
""" Convert from matrix elements describing the discrimination ellipses from Cij (XYZ) to gij (Yxy)"""
SIG = xyz[...,0] + xyz[...,1] + xyz[...,2]
M1 = np.array([SIG, -SIG*xyz[...,0]/xyz[...,1], xyz[...,0]/xyz[...,1]])
M2 = np.array([np.zeros_like(SIG), np.zeros_like(SIG), np.ones_like(SIG)])
M3 = np.array([-SIG, -SIG*(xyz[...,1] + xyz[...,2])/xyz[...,1], xyz[...,2]/xyz[...,1]])
M = np.array((M1,M2,M3))
M = _transpose_02(M) # move stimulus dimension to axis = 0
C = _transpose_02(C) # move stimulus dimension to axis = 0
# convert Cij (XYZ) to gij' (xyY):
AM = np.einsum('ij,kjl->kil', _M_XYZ_TO_PQS, M)
CAM = np.einsum('kij,kjl->kil', C, AM)
# ATCAM = np.einsum('ij,kjl->kil', _M_XYZ_TO_PQS.T, CAM)
# gij = np.einsum('kij,kjl->kil', np.transpose(M,(0,2,1)), ATCAM) # gij = M.T*A.T*C**A*M = (AM).T*C*A*M
gij = np.einsum('kij,kjl->kil', np.transpose(AM,(0,2,1)), CAM) # gij = M.T*A.T*C**A*M = (AM).T*C*A*M
# convert gij' (xyY) to gij (Yxy):
gij = np.roll(np.roll(gij,1,axis=2),1,axis=1)
return gij
def jabz_to_xyz(jabz, ztype='jabz', **kwargs):
"""
Convert Jz,az,bz color coordinates to XYZ tristimulus values.
Args:
:jabz:
| ndarray with Jz,az,bz color coordinates
:ztype:
| 'jabz', optional
| String with requested return:
| Options: 'jabz', 'iabz'
Returns:
:xyz:
| ndarray with tristimulus values
Note:
| 1. :xyz: is assumed to be under D65 viewing conditions! If necessary perform chromatic adaptation!
|
| 2a. Jz represents the 'lightness' relative to a D65 white with luminance = 10000 cd/m²
| (note that Jz that not exactly equal 1 for this high value, but rather for 102900 cd/m2)
| 2b. az, bz represent respectively a red-green and a yellow-blue opponent axis
| (but note that a D65 shows a small offset from (0,0))
Reference:
1. `Safdar, M., Cui, G., Kim,Y. J., and Luo, M. R. (2017).
Perceptually uniform color space for image signals including high dynamic range and wide gamut.
Opt. Express, vol. 25, no. 13, pp. 15131–15151, June, 2017.
<http://www.opticsexpress.org/abstract.cfm?URI=oe-25-13-15131>`_
"""
jabz = np2d(jabz)
# Convert Jz to Iz:
jabz[..., 0] = (jabz[..., 0] + 1.6295499532821566e-11) / (
1 - 0.56 * (1 - (jabz[..., 0] + 1.6295499532821566e-11)))
# Convert Iabz to lmsp:
M = np.linalg.inv(
np.array([[0.5, 0.5, 0], [3.524000, -4.066708, 0.542708],
[0.199076, 1.096799, -1.295875]]))
if len(jabz.shape) == 3:
lmsp = np.einsum('ij,klj->kli', M, jabz)
else:
lmsp = np.einsum('ij,lj->li', M, jabz)
# Convert lmsp to lms:
lms = 10000 * (((3424 / 2**12) - lmsp**(1 / (1.7 * 2523 / 2**5))) /
(((2392 / 2**7) * lmsp**(1 / (1.7 * 2523 / 2**5))) -
(2413 / 2**7)))**(1 / (2610 / (2**14)))
# Convert lms to xyz:
# Setup X',Y',Z' from X,Y,Z transform as matrix:
b = 1.15
g = 0.66
M_to_xyzp = np.array([[b, 0, 1 - b], [1 - g, g, 0], [0, 0, 1]])
# Define X',Y',Z' to L,M,S conversion matrix:
M_to_lms = np.array([[0.41478972, 0.579999, 0.0146480],
[-0.2015100, 1.120649, 0.0531008],
[-0.0166008, 0.264800, 0.6684799]])
# Premultiply M_to_xyzp and M_to_lms and invert:
M = M_to_lms @ M_to_xyzp
M = np.linalg.inv(M)
# Transform L,M,S to X,Y,Z:
if len(jabz.shape) == 3:
xyz = np.einsum('ij,klj->kli', M, lms)
else:
xyz = np.einsum('ij,lj->li', M, lms)
return xyz
def discrimination_hotelling_t2(Yxy1, Yxy2, etype = 'fmc2', ellipsoid = True, Y1 = None, Y2 = None, cspace = 'Yxy'):
"""
Check 'significance' of difference using Hotelling's T2 test on the centers Yxy1 and Yxy2 and their associate FMC-1/2 discrimination ellipses.
Args:
:Yxy1, Yxy2:
| 2D ndarrays with [Y,]x,y coordinate centers.
| If Yxy.shape[-1]==2: Y is added using the value from the Y-input argument.
:etype:
| 'fmc2', optional
| Type of FMC color discrimination equations to use (see references below).
| options: 'fmc1', fmc2'
:Y1, Y2:
| None, optional
| Only affects FMC-2 (see note below).
| If not None: Yi = 10.69 and overrides values in Yxyi.
:ellipsoid:
| True, optional
| If True: return ellipsoids, else return ellipses (only if cspace == 'Yxy')!
:cspace:
| 'Yxy', optional
| Return coefficients for Yxy-ellipses/ellipsoids ('Yxy') or XYZ ellipsoids ('xyz')
Returns:
:p:
| Chi-square based p-value
:T2:
| T2 test statistic (= mahalanobis distance on summed standard error cov. matrices)
Steps:
1. For each center coordinate, the standard error covariance matrix gij^-1 = Si/ni
is determined using the FMC-1 or FMC-2 equations (see refs. 1 & 2).
2. Calculate sum of covariance matrices: SIG = S1/n1 + S2/n2 = gij1^-1 + gij2^-1
3. These are then used in Hotelling's T2 test: T2 = (xy1 - xy2).T*(SIG^-1)*(xy1_xy2)
4. The T2 statistic is then tested against a Chi-square distribution with 2 or 3 degrees of freedom.
References:
1. Chickering, K.D. (1967), Optimization of the MacAdam-Modified 1965 Friele Color-Difference Formula, 57(4):537-541
2. Chickering, K.D. (1971), FMC Color-Difference Formulas: Clarification Concerning Usage, 61(1):118-122
"""
if Yxy1.shape[-1] == 2:
Yxy1 = np.hstack((100*np.ones((Yxy1.shape[0],1)),Yxy1))
if Y1 is not None:
Yxy1[...,0] = Y1
if Yxy2.shape[-1] == 2:
Yxy2 = np.hstack((100*np.ones((Yxy2.shape[0],1)),Yxy2))
if Y2 is not None:
Yxy2[...,0] = Y2
# Get gij matrices (i.e. inverse covariance matrices):
gij1 = get_gij_fmc(Yxy1, etype = etype, ellipsoid=ellipsoid, Y = Y1, cspace = cspace)
gij2 = get_gij_fmc(Yxy2, etype = etype, ellipsoid=ellipsoid, Y = Y2, cspace = cspace)
df = gij1.shape[1] # get degrees of freedom for Chi2
# Calculate T2 statistic:
D12 = (Yxy1[...,(3-df):] - Yxy2[...,(3-df):])
SIG12 = np.linalg.inv(np.linalg.inv(gij1) + np.linalg.inv(gij2))
T2 = np.einsum('ki,ki->k',D12,np.einsum('kij,kj->ki',SIG12,D12))
#T2 = np.atleast_2d(T2).T
# Get p-value:
p = sp.stats.distributions.chi2.sf(T2, df)
return p, T2
def ipt_to_xyz(ipt, cieobs=_CIEOBS, xyzw=None, M=None, **kwargs):
"""
Convert XYZ tristimulus values to IPT color coordinates.
| I: Lightness axis, P, red-green axis, T: yellow-blue axis.
Args:
:ipt:
| ndarray with IPT color coordinates
:xyzw:
| None or ndarray with tristimulus values of white point, optional
| None defaults to xyz of CIE D65 using the :cieobs: observer.
:cieobs:
| luxpy._CIEOBS, optional
| CMF set to use when calculating xyzw for rescaling Mxyz2lms
| (only when not None).
:M: | None, optional
| None defaults to xyz to lms conversion matrix determined by:cieobs:
Returns:
:xyz:
| ndarray with tristimulus values
Note:
:xyz: is assumed to be under D65 viewing conditions! If necessary perform chromatic adaptation !
Reference:
1. `Ebner F, and Fairchild MD (1998).
Development and testing of a color space (IPT) with improved hue uniformity.
In IS&T 6th Color Imaging Conference, (Scottsdale, Arizona, USA), pp. 8–13.
<http://www.ingentaconnect.com/content/ist/cic/1998/00001998/00000001/art00003?crawler=true>`_
"""
ipt = np2d(ipt)
# get M to convert xyz to lms and apply normalization to matrix or input your own:
if M is None:
M = _IPT_M['xyz2lms'][cieobs].copy(
) # matrix conversions from xyz to lms
if xyzw is None:
xyzw = spd_to_xyz(_CIE_ILLUMINANTS['D65'], cieobs=cieobs,
out=1) / 100.0
else:
xyzw = xyzw / 100.0
M = math.normalize_3x3_matrix(M, xyzw)
# convert from ipt to lms':
if len(ipt.shape) == 3:
lmsp = np.einsum('ij,klj->kli', np.linalg.inv(_IPT_M['lms2ipt']), ipt)
else:
lmsp = np.einsum('ij,lj->li', np.linalg.inv(_IPT_M['lms2ipt']), ipt)
# reverse response compression: lms' to lms
lms = lmsp**(1.0 / 0.43)
p = np.where(lmsp < 0.0)
lms[p] = -np.abs(lmsp[p])**(1.0 / 0.43)
# convert from lms to xyz:
if np.ndim(M) == 2:
if len(ipt.shape) == 3:
xyz = np.einsum('ij,klj->kli', np.linalg.inv(M), lms)
else:
xyz = np.einsum('ij,lj->li', np.linalg.inv(M), lms)
else:
if len(
ipt.shape
) == 3: # second dim of lms must match dim of 1st of M and 1st dim of xyzw
xyz = np.concatenate([
np.einsum('ij,klj->kli', np.linalg.inv(M[i]),
lms[:, i:i + 1, :]) for i in range(M.shape[0])
],
axis=1)
else: # first dim of lms must match dim of 1st of M and 1st dim of xyzw
xyz = np.concatenate([
np.einsum('ij,lj->li', np.linalg.inv(M[i]), lms[i:i + 1, :])
for i in range(M.shape[0])
],
axis=0)
#xyz = np.dot(np.linalg.inv(M),lms.T).T
xyz = xyz * 100.0
xyz[np.where(xyz < 0.0)] = 0.0
return xyz
def jabz_to_xyz(jabz, ztype='jabz', use_zcam_parameters=False, **kwargs):
"""
Convert Jz,az,bz color coordinates to XYZ tristimulus values.
Args:
:jabz:
| ndarray with Jz,az,bz color coordinates
:ztype:
| 'jabz', optional
| String with requested return:
| Options: 'jabz', 'iabz'
:use_zcam_parameters:
| False, optional
| ZCAM uses a slightly different values (see notes)
Returns:
:xyz:
| ndarray with tristimulus values
Note:
| 1. :xyz: is assumed to be under D65 viewing conditions! If necessary perform chromatic adaptation!
|
| 2a. Jz represents the 'lightness' relative to a D65 white with luminance = 10000 cd/m²
| (note that Jz that not exactly equal 1 for this high value, but rather for 102900 cd/m2)
| 2b. az, bz represent respectively a red-green and a yellow-blue opponent axis
| (but note that a D65 shows a small offset from (0,0))
| 3. ZCAM: uses a slightly different value for b (=0.7 instead of 0.66)
| and calculates Iz as M' -epsilon (instead L'/2 + M'/2).
Reference:
1. `Safdar, M., Cui, G., Kim,Y. J., and Luo, M. R. (2017).
Perceptually uniform color space for image signals including high dynamic range and wide gamut.
Opt. Express, vol. 25, no. 13, pp. 15131–15151, June, 2017.
<http://www.opticsexpress.org/abstract.cfm?URI=oe-25-13-15131>`_
2. Safdar, M., Hardeberg, J.Y., Luo, M.R. (2021)
"ZCAM, a psychophysical model for colour appearance prediction",
Optics Express.
"""
jabz = np2d(jabz)
# Convert Jz to Iz:
if ztype == 'jabz':
jabz[..., 0] = (jabz[..., 0] + 1.6295499532821566e-11) / (
1 - 0.56 * (1 - (jabz[..., 0] + 1.6295499532821566e-11)))
# Convert Iabz to lmsp:
# Transform L',M',S' to Iabz:
if use_zcam_parameters:
epsilon = 3.70352262101900054e-11
M = _M_LMSP_TO_IAB_ZCAM
else:
epsilon = 0
M = _M_LMSP_TO_IAB_JABZ
jabz[..., 0] += epsilon
if len(jabz.shape) == 3:
lmsp = np.einsum('ij,klj->kli', np.linalg.inv(M), jabz)
else:
lmsp = np.einsum('ij,lj->li', np.linalg.inv(M), jabz)
# Convert lmsp to lms:
c1, c2, c3, n, p = [
_PQ_PARAMETERS[x] for x in sorted(_PQ_PARAMETERS.keys())
]
lms = 10000 * ((c1 - lmsp**(1 / p)) / ((c3 * lmsp**(1 / p)) - c2))**(1 / n)
# Convert lms to xyz:
# Setup X',Y',Z' from X,Y,Z transform as matrix:
b = 1.15
g = 0.66
M_to_xyzp = np.array([[b, 0, 1 - b], [1 - g, g, 0], [0, 0, 1]])
# Premultiply M_to_xyzp and M_to_lms and invert:
M = np.linalg.inv(_M_XYZP_TO_LMS @ M_to_xyzp)
# Transform L,M,S to X,Y,Z:
if len(jabz.shape) == 3:
xyz = np.einsum('ij,klj->kli', M, lms)
else:
xyz = np.einsum('ij,lj->li', M, lms)
return xyz
def xyz_to_jabz(xyz, ztype='jabz', use_zcam_parameters=False, **kwargs):
"""
Convert XYZ tristimulus values to Jz,az,bz color coordinates.
Args:
:xyz:
| ndarray with absolute tristimulus values (Y in cd/m²!)
:ztype:
| 'jabz', optional
| String with requested return:
| Options: 'jabz', 'iabz'
:use_zcam_parameters:
| False, optional
| ZCAM uses a slightly different values (see notes)
Returns:
:jabz:
| ndarray with Jz (or Iz), az, bz color coordinates
Notes:
| 1. :xyz: is assumed to be under D65 viewing conditions! If necessary perform chromatic adaptation!
|
| 2a. Jz represents the 'lightness' relative to a D65 white with luminance = 10000 cd/m²
| (note that Jz that not exactly equal 1 for this high value, but rather for 102900 cd/m2)
| 2b. az, bz represent respectively a red-green and a yellow-blue opponent axis
| (but note that a D65 shows a small offset from (0,0))
| 3. ZCAM: uses a slightly different value for b (=0.7 instead of 0.66)
| and calculates Iz as M' -epsilon (instead L'/2 + M'/2).
Reference:
1. `Safdar, M., Cui, G., Kim,Y. J., and Luo, M. R. (2017).
Perceptually uniform color space for image signals including high dynamic range and wide gamut.
Opt. Express, vol. 25, no. 13, pp. 15131–15151, June 2017.
<http://www.opticsexpress.org/abstract.cfm?URI=oe-25-13-15131>`_
2. Safdar, M., Hardeberg, J.Y., Luo, M.R. (2021)
"ZCAM, a psychophysical model for colour appearance prediction",
Optics Express.
"""
xyz = np2d(xyz)
# Setup X,Y,Z to X',Y',Z' transform as matrix:
b = 1.15
g = 0.66
M_to_xyzp = np.array([[b, 0, 1 - b], [1 - g, g, 0], [0, 0, 1]])
# Premultiply _M_XYZP_TO_LMS and M_to_lms:
M = _M_XYZP_TO_LMS @ M_to_xyzp
# Transform X,Y,Z to L,M,S:
if len(xyz.shape) == 3:
lms = np.einsum('ij,klj->kli', M, xyz)
else:
lms = np.einsum('ij,lj->li', M, xyz)
# response compression: lms to lms'
c1, c2, c3, n, p = [
_PQ_PARAMETERS[x] for x in sorted(_PQ_PARAMETERS.keys())
]
lmsp = ((c1 + c2 * (lms / 10000)**n) / (1 + c3 * (lms / 10000)**n))**p
# Transform L',M',S' to Iabz:
if use_zcam_parameters:
epsilon = 3.70352262101900054e-11
M = _M_LMSP_TO_IAB_ZCAM
else:
epsilon = 0
M = _M_LMSP_TO_IAB_JABZ
if len(lms.shape) == 3:
Iabz = np.einsum('ij,klj->kli', M, lmsp)
else:
Iabz = np.einsum('ij,lj->li', M, lmsp)
Iabz[..., 0] -= epsilon # correct to ensure zero output
# convert Iabz' to Jabz coordinates:
if ztype == 'jabz':
Iabz[..., 0] = ((1 - 0.56) * Iabz[..., 0] /
(1 - 0.56 * Iabz[..., 0])) - 1.6295499532821566e-11
return Iabz
