coloured-spectra 0.0.2

Got a spectra? Use this to find what is the observed colour.

coloured_spectra

Got a spectra? Use this to find the observed colour!

Works great when applied to a single spectra. Using this for hyperspectral though... this Python implementation is too slow to be practical. (Can be magnitudes faster if coded in Julia)

Documentation site: https://yiweimao.github.io/coloured_spectra/

Install

pip install coloured_spectra

Examples

Find the observed colour of a blackbody at 6500 K.

from coloured_spectra import *
import numpy as np
import pickle
import matplotlib.pyplot as plt
import cv2

bb = Blackbody(5778)
bb.plot()

def show_blackbody_colour(T_K):
    bb = Blackbody(T_K,np.linspace(380,750))

    sRGB = spectra2sRGB(bb.λ_nm,bb.B_λT)

    # Due to normalisation choices, the brightness can change depending on the spectra
    # show the colour at max brightness
    HSV = cv2.cvtColor(np.reshape(sRGB,(1,1,3)), cv2.COLOR_RGB2HSV_FULL)
    HSV[0,0,2] = 255
    RGB = cv2.cvtColor(HSV, cv2.COLOR_HSV2RGB_FULL)

    plot_colour(RGB)

This approximates the colour of the Sun with a blackbody temperature of 5778 K.

show_blackbody_colour(5778)

show_blackbody_colour(11000) # what about the star Rigel?

If you have a hyperspectral line, you can colour it in by the observed colour. For this, the invisible UV and NIR is painted white on top of a black background. This look up table spectrum is:

plot_hsv_LUT_spectrum()

# load a hyperspectral line (imaged HgAr lamp)
with open('../wave_cal.pkl','rb') as handle:
    img = np.float32(np.rot90(pickle.load(handle),1))

    # get rid of sensor abnormalities
    img[669,467] = img[670,467]
    img[587,1174] = img[588,1174]

    img /= np.max(img)

plt.imshow(img,cmap='gray',extent=[400,850,0,np.shape(img)[0]])
plt.xlabel("wavelengths (nm)")
plt.ylabel("cross-track")
plt.colorbar()

<matplotlib.colorbar.Colorbar at 0x118865c10>

You can see the emission lines have been coloured.

colour_hyperspectral_line(np.linspace(400,850,2064),img)

100%|██████████| 772/772 [00:41<00:00, 18.53it/s]

You can also return a column of observed colour given a hyperspectral line

hyperspec_line2colour(np.linspace(400,850,2064),img)

100%|██████████| 772/772 [01:07<00:00, 11.46it/s]