1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
|
#!/usr/bin/env python3
# Turns NASA SVS's "Deep Star Maps 2020" Milky Way background (the star map
# without its Hipparcos and Tycho stars, which the sky draws itself) into the
# sRGB JPEG the sky component maps onto the celestial sphere.
#
# curl -O https://svs.gsfc.nasa.gov/vis/a000000/a004800/a004851/milkyway_2020_4k.exr
# python3 tools/sky/build-milkyway.py milkyway_2020_4k.exr public/assets/sky/milkyway-4k.jpg
#
# Needs vips (to read the OpenEXR), numpy and Pillow. The map is plate carrée
# in J2000 right ascension and declination, centred on 0h with r.a. increasing
# to the left; this keeps that layout. Values stay linear in brightness up to
# WHITE (only the heart of the Sagittarius star cloud goes past it), and are
# stored with the sRGB curve so the faint parts keep their detail in 8 bits.
import os
import subprocess
import sys
import tempfile
import numpy as np
from PIL import Image
WHITE = 0.6
def main(src, dst):
width, height = (int(subprocess.check_output(['vipsheader', '-f', f, src], stderr=subprocess.DEVNULL)) for f in ('width', 'height'))
with tempfile.TemporaryDirectory() as tmp:
raw = os.path.join(tmp, 'map.raw')
subprocess.run(['vips', 'rawsave', src, raw], check=True, stderr=subprocess.DEVNULL)
rgb = np.fromfile(raw, dtype=np.float32).reshape(height, width, -1)[..., :3]
x = np.clip(rgb / WHITE, 0.0, 1.0)
srgb = np.where(x <= 0.0031308, 12.92 * x, 1.055 * np.power(x, 1 / 2.4) - 0.055)
# A little noise before rounding keeps the dark gradients from banding.
noise = np.random.default_rng(1).uniform(-0.5, 0.5, srgb.shape)
out = np.clip(np.round(srgb * 255 + noise), 0, 255).astype(np.uint8)
Image.fromarray(out, 'RGB').save(dst, quality=88, optimize=True, progressive=True)
print(f'{width}x{height} -> {dst} ({os.path.getsize(dst) // 1024} KB)')
if __name__ == '__main__':
main(sys.argv[1], sys.argv[2])
|
