cubemap_from_equiangular.py

import os
 
os.environ["OPENCV_IO_MAX_IMAGE_PIXELS"] = pow(2,40).__str__() # Might be needed when loading really big images
#os.environ["OPENCV_IO_ENABLE_OPENEXR"]="1"
 
 
import numpy as np
import cv2
from collections import namedtuple
 
vec3 = namedtuple('vec3', 'x y z')
 
def normalize(v):
    s = 1.0/np.sqrt(v.x*v.x + v.y*v.y + v.z*v.z)
    return vec3(s*v.x, s*v.y, s*v.z)
 
 
img = cv2.imread("Q:\\360\\starmap_2020_64k.png")
 
basises = [
    # [1, V, U]
    [[ 0.0,  0.0, -1.0],   
     [ 0.0,  1.0,  0.0],
     [ 1.0,  0.0,  0.0]],
 
    # [-1, V, U]
    [[ 0.0,  0.0, 1.0],
     [ 0.0,  1.0, 0.0],
     [-1.0,  0.0, 0.0]],
 
    # [U, 1, -V]
    [[ 1.0, 0.0,  0.0],
     [ 0.0, 0.0, -1.0],
     [ 0.0, 1.0,  0.0]],
 
    # [U, -1, V]
    [[ 1.0,  0.0, 0.0],
     [ 0.0,  0.0, 1.0],
     [ 0.0, -1.0, 0.0]],
 
    # [U, V, 1]
    [[ 1.0,  0.0,  0.0],
     [ 0.0,  1.0,  0.0],
     [ 0.0,  0.0,  1.0]],
 
    # [-U, V, -1]
    [[-1.0,  0.0,  0.0],
     [ 0.0,  1.0,  0.0],
     [ 0.0,  0.0, -1.0]]
]
 
s = 16384
#s = 2*16384
t = vec3(( 2.0/s)*(np.outer(np.ones(s),np.arange(s))+0.5) - 1.0,
         ( 2.0/s)*(np.outer(np.arange(s),np.ones(s))+0.5) - 1.0,
         -np.outer(np.ones(s),np.ones(s)))
 
 
print("s={}".format(s))
for f in range(6):
    basis = basises[f]
    d = vec3(basis[0][0]*t.x + basis[0][1]*t.y + basis[0][2]*t.z,
             basis[1][0]*t.x + basis[1][1]*t.y + basis[1][2]*t.z,
             basis[2][0]*t.x + basis[2][1]*t.y + basis[2][2]*t.z)
    d = normalize(d)
 
    u = (img.shape[1]*((0.5/np.pi)*(np.arctan2(d.y, d.x)+np.pi))).astype(int)
    v = (img.shape[0]*((1.0/np.pi)*np.arccos(d.z))).astype(int)
 
    resample = img[v,u];
 
 
    path = "Q:\\360\\moo\\out{}.png".format(f)
    cv2.imwrite(path, resample)
    print("Wrote {}".format(path))