def test_gamma():
source = path('gamma_dalai_lama_gray.jpg')
dalai_lama = snowy.load(source)
snowy.show(dalai_lama)
small = snowy.resize(dalai_lama, height=32)
snowy.save(small, path('small_dalai_lama.png'))
snowy.show(small)
def test_magnification():
i = create_circle(8, 8)
a = snowy.resize(i, 100, 100, snowy.NEAREST)
b = snowy.resize(i, 100, 100, snowy.TRIANGLE)
c = snowy.resize(i, 100, 100, snowy.GAUSSIAN)
e = snowy.resize(i, 100, 100, snowy.MITCHELL)
d = snowy.resize(i, 100, 100, snowy.LANCZOS)
f = snowy.resize(i, 100, 100)
snowy.show(np.hstack([a, b, c, d, e, f]))
def test_minification():
n = snowy.generate_noise(1000, 1000, frequency=5, seed=42)
n = 0.5 + 0.5 * np.sign(n)
a = snowy.resize(n, 100, 100)
b = snowy.resize(n, 100, 100, snowy.MITCHELL)
c = snowy.resize(n, 100, 100, snowy.GAUSSIAN)
d = snowy.resize(n, 100, 100, snowy.NEAREST)
x = [a, b, c, d] + [create_circle(100, 100)]
snowy.show(np.hstack(x))
def test_coords():
h, w = 800, 800
height, width = h, w
# Draw seed image
cyan = np.full([h, w, 3], np.float64([(27, 56, 80)]) / 200)
pink = np.full([h, w, 3], np.float64([175, 111, 127]) / 255)
orange = np.full([h, w, 3], np.float64([239, 159, 95]) / 255)
yellow = np.full([h, w, 3], np.float64([239, 207, 95]) / 255)
colors = np.zeros([h, w, 3])
def max_color(v):
return np.maximum(colors, v)
def sub_color(v):
return colors * (1 - v)
colors = max_color(create_circle(w, h, 0.37, [0.4, 0.5]) * cyan)
colors = max_color(create_circle(w, h, 0.37, [0.6, 0.4]) * cyan)
colors = max_color(create_circle(w, h, 0.27, [0.7, 0.6]) * cyan)
colors = sub_color(create_circle(w, h, 0.35, [0.4, 0.5]))
colors = sub_color(create_circle(w, h, 0.35, [0.6, 0.4]))
colors = sub_color(create_circle(w, h, 0.25, [0.7, 0.6]))
colors = max_color(create_circle(w, h, 0.01, [0.4, 0.5]) * orange)
colors = max_color(create_circle(w, h, 0.01, [0.6, 0.4]) * pink)
colors = max_color(create_circle(w, h, 0.01, [0.7, 0.6]) * yellow)
colors = sn.linearize(colors)
# Create generalized voronoi
luma = sn.reshape(np.sum(colors, 2))
coords = sn.generate_cpcf(luma != 0)
voronoi = sn.dereference_cpcf(colors, coords)
# Warp the voronoi
warpx, warpy = width / 15, height / 15
noise = sn.generate_fBm(width, height, 4, 4, 3)
i, j = np.arange(width, dtype='i2'), np.arange(height, dtype='i2')
coords = np.dstack(np.meshgrid(i, j, sparse=False))
warpx = warpx * np.cos(noise * math.pi * 2)
warpy = warpy * np.sin(noise * math.pi * 2)
coords += np.int16(np.dstack([warpx, warpy]))
coords[:, :, 0] = np.clip(coords[:, :, 0], 0, width - 1)
coords[:, :, 1] = np.clip(coords[:, :, 1], 0, height - 1)
warped = sn.dereference_cpcf(voronoi, coords)
strip = [sn.resize(i, height=256) for i in (colors, voronoi, warped)]
sn.show(sn.hstack(strip))
def test_udf():
c0 = create_circle(200, 200, 0.3)
c1 = create_circle(200, 200, 0.08, 0.8, 0.8)
c0 = np.clip(c0 + c1, 0, 1)
circles = snowy.add_border(c0, value=1)
mask = circles != 0.0
udf = snowy.unitize(snowy.generate_udf(mask))
nx, ny = snowy.gradient(udf)
grad = snowy.unitize(nx + ny)
snowy.show(snowy.hstack([circles, udf, grad]))
def test_tweet():
import snowy as sn, numpy as np
im = sn.generate_noise(2000, 500, 5, seed=2, wrapx=True)
df = sn.generate_sdf(im < 0.0, wrapx=True)
im = 0.5 + 0.5 * np.sign(im) - im
cl = lambda L, U: np.where(np.logical_and(df > L, df < U), -im, 0)
im += cl(20, 30) + cl(60, 70) + cl(100, 110)
sn.show(sn.resize(im, height=100, wrapx=True))
sn.show(sn.resize(np.hstack([im, im]), height=200, wrapx=True))
def draw_quad():
verts = np.array([[-0.67608007, 0.38439575, 3.70544936, 0., 0. ],
[-0.10726266, 0.38439575, 2.57742041, 1., 0. ],
[-0.10726266, -0.96069041, 2.57742041, 1., 1. ],
[-0.67608007, -0.96069041, 3.70544936, 0., 1. ]])
texture = snowy.load(qualify('../tests/texture.png'))
target = np.full((1080, 1920, 4), (0.54, 0.54, 0.78, 1.00),
dtype=np.float32)
snowy.draw_polygon(target, texture, verts)
target = snowy.resize(target[400:770, 700:1000], height = 256)
texture = snowy.resize(texture, height = 256)
quad = snowy.hstack([texture, target])
snowy.export(quad, qualify('quad.png'))
snowy.show(quad)
def test_normals():
isle = create_island(10)
height, width, nchan = isle.shape
occlusion = np.empty([height, width, 1])
seconds = timeit.timeit(
lambda: np.copyto(occlusion, sn.compute_skylight(isle)), number=1)
print(f'\ncompute_skylight took {seconds} seconds')
normals = np.empty([height - 1, width - 1, 3])
seconds = timeit.timeit(
lambda: np.copyto(normals, sn.compute_normals(isle)), number=1)
print(f'\ncompute_normals took {seconds} seconds')
normals = sn.resize(normals, 750, 512)
# Flatten the normals according to landmass versus sea.
normals += np.float64([0, 0, 100]) * np.where(isle < 0.0, 1.0, 0.005)
normals /= sn.reshape(np.sqrt(np.sum(normals * normals, 2)))
# Compute the lambertian diffuse factor
lightdir = np.float64([0.2, -0.2, 1])
lightdir /= np.linalg.norm(lightdir)
df = np.clip(np.sum(normals * lightdir, 2), 0, 1)
df = sn.reshape(df)
df *= occlusion
# Apply color LUT
gradient_image = sn.resize(sn.load(path('terrain.png')),
width=1024)[:, :, :3]
def applyColorGradient(elevation):
xvals = np.arange(1024)
yvals = gradient_image[0]
apply_lut = interpolate.interp1d(xvals, yvals, axis=0)
el = np.clip(1023 * elevation, 0, 1023)
return apply_lut(sn.unshape(el))
albedo = applyColorGradient(isle * 0.5 + 0.5)
albedo *= df
# Visualize the lighting layers
normals = 0.5 * (normals + 1.0)
isle = np.dstack([isle, isle, isle])
occlusion = np.dstack([occlusion, occlusion, occlusion])
df = np.dstack([df, df, df])
island_strip = sn.resize(sn.hstack([occlusion, normals, df, albedo]),
height=256)
sn.save(island_strip, 'docs/island_strip.png')
sn.show(island_strip)
def test_gdf():
"This is a (failed) effort to create a smoother distance field."
c0 = create_circle(200, 200, 0.3)
c1 = create_circle(200, 200, 0.08, 0.8, 0.8)
c0 = np.clip(c0 + c1, 0, 1)
circles = snowy.add_border(c0, value=1)
circles = np.clip(snowy.blur(circles, radius=2), 0, 1)
circles = np.clip(snowy.blur(circles, radius=2), 0, 1)
source = (1.0 - circles) * 2000.0
gdf = np.sqrt(snowy.generate_gdf(source))
gdf = snowy.unitize(gdf)
nx, ny = snowy.gradient(gdf)
grad = snowy.unitize(nx + ny)
snowy.show(snowy.hstack([circles, gdf, grad]))
def test_range():
source = path('../docs/ground.jpg')
ground = snowy.load(source)
assert np.amin(ground) >= 0 and np.amax(ground) <= 1
with tempfile.NamedTemporaryFile() as fp:
target = fp.name + '.png'
snowy.save(ground, target)
show_filename(target)
show_filename(source)
show_array(ground)
blurred = snowy.blur(ground, radius=10)
snowy.show(blurred)
def test_tileable():
n = snowy.generate_noise(200, 400, frequency=4, seed=42, wrapx=True)
n = 0.5 + 0.5 * np.sign(n) - n
n = np.hstack([n, n])
gold = snowy.resize(n, 200, 200)
n = snowy.generate_noise(20, 40, frequency=4, seed=42, wrapx=True)
n = 0.5 + 0.5 * np.sign(n) - n
n = snowy.resize(n, 100, 200)
bad = np.hstack([n, n])
n = snowy.generate_noise(20, 40, frequency=4, seed=42, wrapx=True)
n = 0.5 + 0.5 * np.sign(n) - n
n = snowy.resize(n, 100, 200, wrapx=True)
good = np.hstack([n, n])
snowy.show(snowy.hstack([gold, bad, good], 2, .7))
def test_tileable_distance():
c0 = create_circle(400, 200, 0.3)
c1 = create_circle(400, 200, 0.08, 0.8, 0.8)
circles = np.clip(c0 + c1, 0, 1)
mask = circles != 0.0
sdf = snowy.unitize(snowy.generate_sdf(mask, wrapx=True, wrapy=True))
nx, ny = snowy.gradient(sdf)
grad = snowy.unitize(nx + ny)
stack2 = np.hstack([sdf, sdf, grad, grad])
snowy.show(snowy.resize(np.vstack([stack2, stack2]), 600, 200))
get_mask = lambda L, U: np.logical_and(sdf > L, sdf < U)
get_contour = lambda L, U: np.where(get_mask(L, U), sdf, 0)
sdf -= get_contour(.20, .25)
sdf -= get_contour(.60, .65)
sdf -= get_contour(.90, .95)
snowy.show(snowy.resize(np.hstack([sdf, sdf, sdf, sdf]), height=300))
def test_luminance():
source = sn.load('tests/sobel_input.png')[:, :, :3]
L = rgb2gray(source)
skresult = np.dstack([L, L, L])
small_skresult = sn.resize(skresult, width=256)
L = sn.rgb_to_luminance(source)
snresult = np.dstack([L, L, L])
small_snresult = sn.resize(snresult, width=256)
L = skimage_sobel(source)
sksobel = np.dstack([L, L, L])
small_sksobel = sn.resize(sksobel, width=256)
L = sn.rgb_to_luminance(source)
L = sn.compute_sobel(L)
snsobel = np.dstack([L, L, L])
small_snsobel = sn.resize(snsobel, width=256)
sn.show(
np.hstack(
[small_skresult, small_snresult, small_sksobel, small_snsobel]))
def test_thick():
source = sn.load('tests/sobel_input.png')[:, :, :3]
small_source = sn.resize(source, width=256)
blurred = sn.blur(source, radius=2)
small_blurred = sn.resize(blurred, width=256)
L = skimage_sobel(blurred)
sksobel = np.dstack([L, L, L])
small_sksobel = sn.resize(sksobel, width=256)
L = sn.rgb_to_luminance(blurred)
L = sn.compute_sobel(L)
snsobel = np.dstack([L, L, L])
small_snsobel = sn.resize(snsobel, width=256)
small_sksobel = np.clip(1 - small_sksobel * 40, 0, 1)
small_snsobel = np.clip(1 - small_snsobel * 40, 0, 1)
strip = np.hstack([
small_blurred, small_source * small_sksobel,
small_source * small_snsobel
])
sn.show(strip)
def test_draw_quad2():
target = np.full((1080, 1920, 4), (0, 0, 0, 0), dtype=np.float32)
texture = snowy.load('tests/texture.png')
# These are in NDC so they span -W to +W
vertices = np.array([[-0.67608007, 0.38439575, 1.7601049, 3.70544936],
[-0.10726266, 0.38439575, 0.60928749, 2.57742041],
[-0.10726266, -0.96069041, 0.60928749, 2.57742041],
[-0.67608007, -0.96069041, 1.7601049, 3.70544936]])
texcoords = np.array([[0., 0.], [1., 0.], [1., 1.], [0., 1.]])
x, y, w = vertices[:, 0], vertices[:, 1], vertices[:, 3]
u, v = texcoords[:, 0], texcoords[:, 1]
vertices = np.transpose(np.vstack([x, y, w, u, v]))
print(vertices)
snowy.draw_polygon(target, texture, vertices)
overlay = snowy.load('tests/overlay.png')
im = snowy.compose(target, overlay)[400:770, 600:900]
target = snowy.resize(im, height=100)
snowy.show(target)
gibbons = snowy.load(qualify('gibbons.jpg'))
gibbons = snowy.resize(gibbons, width=gibbons.shape[1] // 5)
gibbons90 = snowy.rotate(gibbons, 90)
gibbons180 = snowy.rotate(gibbons, 180)
gibbons270 = snowy.rotate(gibbons, 270)
hflipped = snowy.hflip(gibbons)
vflipped = snowy.vflip(gibbons)
snowy.export(snowy.hstack([gibbons, gibbons180, vflipped],
border_width=4, border_value=[0.5,0,0]), qualify("xforms.png"))
# Test noise generation
n = snowy.generate_noise(100, 100, frequency=4, seed=42, wrapx=True)
n = np.hstack([n, n])
n = 0.5 + 0.5 * n
snowy.show(n)
snowy.export(n, qualify('noise.png'))
# First try minifying grayscale
gibbons = snowy.load(qualify('snowy.jpg'))
gibbons = np.swapaxes(gibbons, 0, 2)
gibbons = np.swapaxes(gibbons[0], 0, 1)
gibbons = snowy.reshape(gibbons)
source = snowy.resize(gibbons, height=200)
blurry = snowy.blur(source, radius=4.0)
diptych_filename = qualify('diptych.png')
snowy.export(snowy.hstack([source, blurry]), diptych_filename)
optimize(diptych_filename)
snowy.show(diptych_filename)
# 1. Create falloff shape. import snowy import numpy as np from functools import reduce from scipy import interpolate width, height = 768, 256 x, y = np.linspace(-1, 1, width), np.linspace(-1, 1, height) u, v = np.meshgrid(x, y, sparse=True) falloff = np.clip(1 - (u * u + v * v), 0, 1) falloff = snowy.reshape(falloff / 2) snowy.show(falloff) # 2. Add layers of gradient noise and scale with falloff. noise = snowy.generate_noise noise = [noise(width, height, 6 * 2**f, int(f)) * 1 / 2**f for f in range(4)] noise = reduce(lambda x, y: x + y, noise) elevation = falloff * (falloff / 2 + noise) elevation = snowy.generate_udf(elevation < 0.1) elevation /= np.amax(elevation) snowy.show(elevation) # 3. Compute ambient occlusion. occlusion = snowy.compute_skylight(elevation) snowy.show(occlusion) # 4. Generate normal map.
def show(im):
snowy.show(snowy.resize(im, height=100, filter=None))
def test_noise_smoothness():
noise = 0.5 + 0.5 * snowy.generate_noise(300, 150, 4, seed=42)
grad = snowy.gradient(noise)
grad = grad[0] + grad[1]
grad = snowy.unitize(grad)
snowy.show(grad)
def test_solid():
gray = np.ones([100, 100, 4]) / 2
snowy.show(gray)
def clumpy(cmd):
print(f"clumpy {cmd}")
result = system(f".release/clumpy {cmd}")
if result: raise Exception("clumpy failed")
friction = 0.9
spacing = 20
step_size = 2.5
kernel_size = 5
decay = 0.99
nframes = 400
res = 4000, 2000
scalex = 2
scaley = 5
dim = "x".join(map(str, res))
clumpy(f"pendulum_phase {dim} {friction} {scalex} {scaley} field.npy")
clumpy(f"bridson_points {dim} {spacing} 0 pts.npy")
clumpy(
f"advect_points pts.npy field.npy {step_size} {kernel_size} {decay} {nframes} phase.npy"
)
im = snowy.reshape(1.0 - np.load("000phase.npy") / 255.0)
im = snowy.resize(im, 500, 250)
snowy.export(im, "extras/example6.png")
snowy.show(im)
system("rm *.npy")
print("Generated extras/example6.png")
import snowy
source = snowy.open('poodle.png')
source = snowy.resize(source, height=200)
blurry = snowy.blur(source, radius=4.0)
snowy.save(snowy.hstack([source, blurry]), 'diptych.png')
# This snippet does a resize, then a blur, then horizontally concatenates the two images
parrot = snowy.load('parrot.png')
height, width = parrot.shape[:2]
nearest = snowy.resize(parrot, width * 6, filter=snowy.NEAREST)
mitchell = snowy.resize(parrot, width * 6)
snowy.show(snowy.hstack([nearest, mitchell]))
# This snippet first magnifies an image using a nearest-neighbor filter, then using the default Mitchell filter.
parrot = snowy.load('parrot.png')
height, width = parrot.shape[:2]
nearest = snowy.resize(parrot, width * 6, filter=snowy.NEAREST)
mitchell = snowy.resize(parrot, width * 6)
snowy.show(snowy.hstack([nearest, mitchell]))
