Implementation of the Helmholtz-Kohlrausch effect

[geek9012]

Hello, I have been trying to use the library to implement the HelmholtzKohlrausch effect.
My two references are:

1. https://github.com/ilia3101/HKE/blob/main/hke.py
2. https://colour.readthedocs.io/en/latest/generated/colour.HelmholtzKohlrausch_effect_object_Nayatani1997.html

So, after I applied the corrections, both images still looked the same.
Before correction:

After correction:

Plotted difference:

My python code is attached as a text file.

Any help is appreciated.
Kind regards

[KelSolaar]

Hello @geek9012,

I cannot see the code on Github, would it be possible to paste it in a comment please?

Cheers,

Thomas

[geek9012]

import colour
from colour.appearance.hke import*
import matplotlib.pyplot as plt
import matplotlib.patches as mpathes
import numpy as np
import cv2 as cv

#the function below taken from https://github.com/ilia3101/HKE/blob/main/hke.py
def plot_patches(FileName, PatchesXYZ, BackgroundColour):
    #dont know what this does
    fig,ax = plt.subplots()
    #background grey
    background_rect = mpathes.Rectangle([0.0,0.0],0.3+len(patches)*0.3,0.32,color=BackgroundColour)
    ax.add_patch(background_rect)
    #Draw all patches
    patch_count = 0
    print('RGBs of patches:')
    for patch in PatchesXYZ:
        rgb = colour.XYZ_to_sRGB(patch)
        print(rgb)
        ax.add_patch(mpathes.Rectangle([0.2+patch_count*0.3, 0.06], 0.2, 0.2, color=colour.notation.RGB_to_HEX(rgb)))
        patch_count += 1
    # Save image
    plt.axis('scaled')
    plt.axis('off')
    plt.tight_layout()
    plt.savefig(FileName, dpi=500,bbox_inches='tight',pad_inches=0,transparent=True)

#Luminance of background, ref: https://github.com/ilia3101/HKE/blob/main/hke.py
Y_value = 0.74
Y_sRGB = colour.models.eotf_inverse_sRGB(Y_value)

L_adapting = 65

#Inspired by the image on https://en.wikipedia.org/wiki/Helmholtz%E2%80%93Kohlrausch_effect
# ***Patch colours specified in sRGB here
patches = [
    [248, 0, 0],
    [118, 131, 0],
    [0, 142, 92],
    [0, 130, 197],
    [212, 0, 234],
]

#normalise to 0-1 and linearise
for p in patches:
    for i in range (0,3):
        p[i] /= 255.0
        
patches_XYZ = []
for patch in patches:
    in_XYZ = colour.sRGB_to_XYZ(patch, apply_cctf_decoding=True)
    Y_mul = (L_adapting/in_XYZ[1])
    for i in range(0,3): in_XYZ[i] *= Y_mul #Normalise Y value
    patches_XYZ.append(in_XYZ)
    
white = colour.xy_to_Luv_uv(colour.temperature.CCT_to_xy_CIE_D(6504))
colours = colour.XYZ_to_xy(patches_XYZ)

adaptation_factors = HelmholtzKohlrausch_effect_object_Nayatani1997(  # doctest: +ELLIPSIS
      colour.xy_to_Luv_uv(colours), white, L_adapting
     )

colours_xyY = np.column_stack((colours, np.full((colours.shape[0], 1), L_adapting)))
# Divide Y by adaptation factors
colours_Y = L_adapting* adaptation_factors
# Create xyY array with adjusted Y values
colours_xyY_adapted = np.column_stack((colours,colours_Y[:, None]))
colours_XYZ_adapted = colour.xyY_to_XYZ(colours_xyY_adapted)

grey_colour = colour.notation.RGB_to_HEX([Y_sRGB,Y_sRGB,Y_sRGB])
print('Grey:',[Y_sRGB,Y_sRGB,Y_sRGB])
plot_patches("images/equal_Y1.png", patches_XYZ, grey_colour)

plot_patches("images/equal_Y2.png", colours_XYZ_adapted, grey_colour)

#Plot the Difference:
# Load the images
image1 = cv.imread("images/equal_Y1.png")
image2 = cv.imread("images/equal_Y2.png")

# Convert images to grayscale (assuming they're color images)
gray1 = cv.cvtColor(image1, cv.COLOR_BGR2GRAY)
gray2 = cv.cvtColor(image2, cv.COLOR_BGR2GRAY)

# Compute the absolute difference
diff = cv.absdiff(gray1, gray2)

# Plot the difference
plt.figure(figsize=(10, 5))

plt.subplot(1, 3, 1)
plt.imshow(gray1, cmap='gray')
plt.title('Image 1')

plt.subplot(1, 3, 2)
plt.imshow(gray2, cmap='gray')
plt.title('Image 2')

plt.subplot(1, 3, 3)
plt.imshow(diff, cmap='gray')
plt.title('Difference')

sum(sum(diff))

[geek9012]

I found one mistake in my code. I used the adaptation field parameter wrong. I used L_adapting = 65 which should not be there. So, I have this corrected version of my code:

import colour
from colour.appearance.hke import*
import matplotlib.pyplot as plt
import matplotlib.patches as mpathes
import numpy as np
import cv2 as cv

def plot_patches(FileName, PatchesXYZ, BackgroundColour):
    #dont know what this does
    fig,ax = plt.subplots()
    #background grey
    background_rect = mpathes.Rectangle([0.0,0.0],0.3+len(patches)*0.3,0.32,color=BackgroundColour)
    ax.add_patch(background_rect)
    #Draw all patches
    patch_count = 0
    print('RGBs of patches:')
    for patch in PatchesXYZ:
        rgb = colour.XYZ_to_sRGB(patch)
        print(rgb)
        ax.add_patch(mpathes.Rectangle([0.2+patch_count*0.3, 0.06], 0.2, 0.2, color=colour.notation.RGB_to_HEX(rgb)))
        patch_count += 1
    # Save image
    plt.axis('scaled')
    plt.axis('off')
    plt.tight_layout()
    plt.savefig(FileName, dpi=500,bbox_inches='tight',pad_inches=0,transparent=True)

#Luminance of background, ref: https://github.com/ilia3101/HKE/blob/main/hke.py
Y_value = 0.14
Y_sRGB = colour.models.eotf_inverse_sRGB(Y_value)


#Inspired by the image on https://en.wikipedia.org/wiki/Helmholtz%E2%80%93Kohlrausch_effect
# ***Patch colours specified in sRGB here
patches = [
    [248, 0, 0],
    [118, 131, 0],
    [0, 142, 92],
    [0, 130, 197],
    [212, 0, 234],
]

#normalise to 0-1 and linearise
for p in patches:
    for i in range (0,3):
        p[i] /= 255.0
        
patches_XYZ = []
for patch in patches:
    in_XYZ = colour.sRGB_to_XYZ(patch, apply_cctf_decoding=True)
    Y_mul = (Y_value/in_XYZ[1])
    for i in range(0,3): in_XYZ[i] *= Y_mul #Normalise Y value
    patches_XYZ.append(in_XYZ)
    
white = colour.xy_to_Luv_uv(colour.temperature.CCT_to_xy_CIE_D(6504))
colours = colour.XYZ_to_xy(patches_XYZ)

adaptation_factors = HelmholtzKohlrausch_effect_object_Nayatani1997(  # doctest: +ELLIPSIS
      colour.xy_to_Luv_uv(colours), white, Y_value
     )

colours_xyY = np.column_stack((colours, np.full((colours.shape[0], 1), Y_value)))
# Divide Y by adaptation factors
colours_Y = Y_value / adaptation_factors
# Create xyY array with adjusted Y values
colours_xyY_adapted = np.column_stack((colours,colours_Y[:, None]))
colours_XYZ_adapted = colour.xyY_to_XYZ(colours_xyY_adapted)

grey_colour = colour.notation.RGB_to_HEX([Y_sRGB,Y_sRGB,Y_sRGB])
print('Grey:',[Y_sRGB,Y_sRGB,Y_sRGB])
plot_patches("images/equal_Y1.png", patches_XYZ, grey_colour)

plot_patches("images/equal_Y2.png", colours_XYZ_adapted, grey_colour)

#Plot the Difference:
# Load the images
image1 = cv.imread("images/equal_Y1.png")
image2 = cv.imread("images/equal_Y2.png")

# Convert images to grayscale (assuming they're color images)
gray1 = cv.cvtColor(image1, cv.COLOR_BGR2GRAY)
gray2 = cv.cvtColor(image2, cv.COLOR_BGR2GRAY)

# Compute the absolute difference
diff = cv.absdiff(gray1, gray2)

# Plot the difference
plt.figure(figsize=(10, 5))

plt.subplot(1, 3, 1)
plt.imshow(gray1, cmap='gray')
plt.title('Image 1')

plt.subplot(1, 3, 2)
plt.imshow(gray2, cmap='gray')
plt.title('Image 2')

plt.subplot(1, 3, 3)
plt.imshow(diff, cmap='gray')
plt.title('Difference')

sum(sum(diff))

Now, the images look better and I get a difference image as well.

Original image:

Corrected image:

Difference:

But the corrected image still looks a bit incorrect. The red is still a bit brighter, I feel. Do you have any suggestions to correct it or do you see anything wrong?

Kind regards

[KelSolaar]

They look a bit too bright indeed, we have an example here: https://github.com/colour-science/colour/blob/develop/colour/examples/appearance/examples_hke.py

"""Showcases Helmholtz—Kohlrausch effect estimation computations."""

import colour
from colour.plotting import colour_style, plot_multi_colour_swatches
from colour.utilities import message_box

wp = colour.xy_to_Luv_uv([0.31271, 0.32902])

average_luminance = 0.14

swatches = [
    [0.45079660, 0.52288689],
    [0.19124902, 0.55444488],
    [0.13128455, 0.51210591],
    [0.14889223, 0.37091478],
    [0.28992574, 0.30964533],
]
swatches_XYZ = []
for patch in swatches:
    in_XYZ = colour.Luv_to_XYZ(colour.uv_to_Luv(patch))
    swatches_XYZ.append(in_XYZ * (average_luminance / in_XYZ[1]))

# Adapting Luminance, 250 cd/m^2 represents a typical modern computer
# display peak luminance.
L_a = 250 * average_luminance

bg_grey = colour.xy_to_XYZ(colour.Luv_uv_to_xy(wp)) * average_luminance

swatches_normal = []
swatches_VCC = []
swatches_VAC = []

for i in range(len(swatches)):
    VCC = colour.HelmholtzKohlrausch_effect_luminous_Nayatani1997(
        swatches[i], wp, L_a, method="VCC"
    )
    VAC = colour.HelmholtzKohlrausch_effect_luminous_Nayatani1997(
        swatches[i], wp, L_a, method="VAC"
    )

    swatches_normal.append(colour.XYZ_to_sRGB(bg_grey))
    swatches_normal.append(colour.XYZ_to_sRGB(swatches_XYZ[i]))

    swatches_VCC.append(colour.XYZ_to_sRGB(bg_grey))
    swatches_VCC.append(colour.XYZ_to_sRGB(swatches_XYZ[i] / VCC))

    swatches_VAC.append(colour.XYZ_to_sRGB(bg_grey * VAC))
    swatches_VAC.append(colour.XYZ_to_sRGB(swatches_XYZ[i]))

colour_style()

message_box(
    "Plotting swatches with the same luminance (Y).\n"
    "The Helmholtz—Kohlrausch effect will be very noticeable."
)
plot_multi_colour_swatches(swatches_normal, compare_swatches="stacked")

message_box(
    "Plotting HKE-compensated swatches with VCC method.\n"
    "The Helmholtz—Kohlrausch effect has been compensated using VCC"
    "(variable chromatic colour) method."
)
plot_multi_colour_swatches(swatches_VCC, compare_swatches="stacked")

message_box(
    "Plotting HKE-compensated swatches with VAC method.\n"
    "The Helmholtz—Kohlrausch effect has been compensated for using VAC"
    "(variable achromatic colour) method."
)
plot_multi_colour_swatches(swatches_VAC, compare_swatches="stacked")