def computeInitialNormal(self):
if self._N_lumo.shape != self._N_b_smooth.shape:
return
self._N0_b_32F = normalizeImage(
bumpMapping(np.array(self._N_lumo), self._N_b_smooth))
self._N0_d_32F = normalizeImage(
bumpMapping(np.array(self._N_lumo), self._N_d_smooth))
def _computeInitialDetailNormal(self):
bump_scale = self._parameters["bumpScale"].value()
self._N_b[:, :, :2] *= bump_scale
self._N_b = normalizeImage(self._N_b, th=1.0)
self._N_d[:, :, :2] *= bump_scale
self._N_d = normalizeImage(self._N_d, th=1.0)
self._view.render(normalToColor(self._N_b))
def _computeInitialDetailNormal(self):
bump_scale = self._parameters["bumpScale"].value()
self._N_b[:, :, :2] *= bump_scale
self._N_b = normalizeImage(self._N_b, th=1.0)
self._N_d[:, :, :2] *= bump_scale
self._N_d = normalizeImage(self._N_d, th=1.0)
self._view.render(normalToColor(self._N_b))
def computeGradientNormals(D_32F, sigma=1.0):
h, w = D_32F.shape
gx = cv2.Sobel(D_32F, cv2.CV_64F, 1, 0, ksize=1)
gx = cv2.GaussianBlur(gx, (0, 0), sigma)
gy = cv2.Sobel(D_32F, cv2.CV_64F, 0, 1, ksize=1)
gy = cv2.GaussianBlur(gy, (0, 0), sigma)
g_max = max(np.max(gx), np.max(gy))
g_scale = 100.0 / g_max
T_32F = np.zeros((h, w, 3), dtype=np.float32)
T_32F[:, :, 0] = 1.0
T_32F[:, :, 2] = g_scale * gx
B_32F = np.zeros((h, w, 3), dtype=np.float32)
B_32F[:, :, 1] = 1.0
B_32F[:, :, 2] = -g_scale * gy
T_flat = T_32F.reshape(-1, 3)
B_flat = B_32F.reshape(-1, 3)
N_flat = np.cross(T_flat, B_flat)
N_32F = N_flat.reshape(h, w, 3)
N_32F = normalizeImage(N_32F)
return N_32F
def computeGradientNormals(D_32F, sigma=1.0):
h, w = D_32F.shape
gx = cv2.Sobel(D_32F, cv2.CV_64F, 1, 0, ksize=1)
gx = cv2.GaussianBlur(gx, (0, 0), sigma)
gy = cv2.Sobel(D_32F, cv2.CV_64F, 0, 1, ksize=1)
gy = cv2.GaussianBlur(gy, (0, 0), sigma)
g_max = max(np.max(gx), np.max(gy))
g_scale = 100.0 / g_max
T_32F = np.zeros((h, w, 3), dtype=np.float32)
T_32F[:, :, 0] = 1.0
T_32F[:, :, 2] = g_scale * gx
B_32F = np.zeros((h, w, 3), dtype=np.float32)
B_32F[:, :, 1] = 1.0
B_32F[:, :, 2] = -g_scale * gy
T_flat = T_32F.reshape(-1, 3)
B_flat = B_32F.reshape(-1, 3)
N_flat = np.cross(T_flat, B_flat)
N_32F = N_flat.reshape(h, w, 3)
N_32F = normalizeImage(N_32F)
return N_32F
def depthToNormal(D_32F):
h, w = D_32F.shape
gx = cv2.Sobel(D_32F, cv2.CV_64F, 1, 0, ksize=1)
gy = cv2.Sobel(D_32F, cv2.CV_64F, 0, 1, ksize=1)
N_32F = np.zeros((h, w, 3), dtype=np.float32)
N_32F[:, :, 0] = -gx
N_32F[:, :, 1] = gy
N_32F[:, :, 2] = 2
N_32F = normalizeImage(N_32F)
return N_32F
def preProcess(N0_32F, A_8U):
foreground = A_8U > 0.5 * np.max(A_8U)
background = A_8U == 0
N_32F = np.array(N0_32F)
N_32F[background, :] = np.array([0.0, 0.0, 1.0])
sigma = 5.0
for i in xrange(5):
N_32F = cv2.GaussianBlur(N_32F, (0, 0), sigma)
N_32F[foreground, :] = N0_32F[foreground, :]
#N_32F[background, :] = np.array([0.0, 0.0, 1.0])
#N_32F[background, :] = np.array([0.0, 0.0, 1.0])
N_32F = normalizeImage(N_32F)
return N_32F
def preProcess(N0_32F, A_8U):
foreground = A_8U > 0.5 * np.max(A_8U)
background = A_8U == 0
N_32F = np.array(N0_32F)
N_32F[background, :] = np.array([0.0, 0.0, 1.0])
sigma = 5.0
for i in xrange(5):
N_32F = cv2.GaussianBlur(N_32F, (0, 0), sigma)
N_32F[foreground, :] = N0_32F[foreground, :]
#N_32F[background, :] = np.array([0.0, 0.0, 1.0])
#N_32F[background, :] = np.array([0.0, 0.0, 1.0])
N_32F = normalizeImage(N_32F)
return N_32F
def bumpNormal(D_32F, scale=1.0, sigma=1.0):
gx = cv2.Sobel(D_32F, cv2.CV_64F, 1, 0, ksize=1)
gx = cv2.GaussianBlur(gx, (0, 0), sigma)
gy = -cv2.Sobel(D_32F, cv2.CV_64F, 0, 1, ksize=1)
gy = cv2.GaussianBlur(gy, (0, 0), sigma)
h, w = D_32F.shape[:2]
N_32F = np.zeros((h, w, 3), dtype=np.float32)
N_32F[:, :, 0] = -scale * gx
N_32F[:, :, 1] = -scale * gy
N_32F[:, :, 2] = 1.0
N_32F = normalizeImage(N_32F, th=1.0)
return N_32F
def bumpNormal(D_32F, scale=1.0, sigma=1.0):
gx = cv2.Sobel(D_32F, cv2.CV_64F, 1, 0, ksize=1)
gx = cv2.GaussianBlur(gx, (0, 0), sigma)
gy = -cv2.Sobel(D_32F, cv2.CV_64F, 0, 1, ksize=1)
gy = cv2.GaussianBlur(gy, (0, 0), sigma)
h, w = D_32F.shape[:2]
N_32F = np.zeros((h, w, 3), dtype=np.float32)
N_32F[:, :, 0] = -scale * gx
N_32F[:, :, 1] = -scale * gy
N_32F[:, :, 2] = 1.0
N_32F = normalizeImage(N_32F, th=1.0)
return N_32F
def _interpolateNormalAMG(self, N0_32F, W_32F, A_8U):
h, w = N0_32F.shape[:2]
A_c, b_c = normalConstraints(W_32F, N0_32F)
A_8U = None
if self._image.shape[2] == 4:
A_8U = to8U(alpha(self._image))
A_sil, b_sil = silhouetteConstraints(A_8U)
A_L = laplacianMatrix((h, w))
A = 10.0 * A_c + A_L + A_sil
b = 10.0 * b_c + b_sil
N_32F = amg_solver.solve(A, b).reshape(h, w, 3)
N_32F = normalizeImage(N_32F)
return N_32F
def normalSphere(h=256, w=256):
N_32F = np.zeros((h, w, 3))
A_32F = np.zeros((h, w))
for y in xrange(h):
N_32F[y, :, 0] = np.linspace(-1.05, 1.05, w)
for x in xrange(w):
N_32F[:, x, 1] = np.linspace(1.05, -1.05, w)
r_xy = N_32F[:, :, 0]**2 + N_32F[:, :, 1]**2
N_32F[r_xy < 1.0, 2] = np.sqrt(1.0 - r_xy[r_xy < 1.0])
N_32F[r_xy > 1.0, 2] = 0.0
N_32F = normalizeImage(N_32F)
A_32F[r_xy < 1.0] = 1.0 - r_xy[r_xy < 1.0]**100
A_32F = cv2.bilateralFilter(np.float32(A_32F), 0, 0.1, 2)
return N_32F, A_32F
def _computeLumoNormal(self):
A_8U = self._A_8U
if A_8U is None:
return
h, w = A_8U.shape[:2]
A_c, b_c = amg_constraints.silhouetteConstraints(A_8U)
A_L = amg_constraints.laplacianMatrix((h, w))
A = 3.0 * A_c + A_L
b = 3.0 * b_c
N_32F = amg_solver.solve(A, b).reshape(h, w, 3)
N_32F = computeNz(N_32F.reshape(-1, 3)).reshape(h, w, 3)
N_32F = normalizeImage(N_32F)
self._N_lumo = np.array(N_32F)
def _computeLumoNormal(self):
A_8U = self._A_8U
if A_8U is None:
return
h, w = A_8U.shape[:2]
A_c, b_c = amg_constraints.silhouetteConstraints(A_8U)
A_L = amg_constraints.laplacianMatrix((h, w))
A = 3.0 * A_c + A_L
b = 3.0 * b_c
N_32F = amg_solver.solve(A, b).reshape(h, w, 3)
N_32F = computeNz(N_32F.reshape(-1, 3)).reshape(h, w, 3)
N_32F = normalizeImage(N_32F)
self._N_lumo = np.array(N_32F)
def _computeDetailNormal(self, N0_32F):
h, w = N0_32F.shape[:2]
W_32F = np.zeros((h, w))
# sigma_d = 2.0 * np.max(N0_32F[:, :, 2])
# W_32F = 1.0 - np.exp( - (N0_32F[:, :, 2] ** 2) / (sigma_d ** 2))
W_32F = 1.0 - N0_32F[:, :, 2]
W_32F *= 1.0 / np.max(W_32F)
W_32F = W_32F ** 1.5
A_c, b_c = amg_constraints.normalConstraints(W_32F, N0_32F)
A_L = amg_constraints.laplacianMatrix((h, w))
lambda_d = 2.0
A = A_c + lambda_d * A_L
b = b_c
N_32F = amg_solver.solve(A, b).reshape(h, w, 3)
N_32F = computeNz(N_32F.reshape(-1, 3)).reshape(h, w, 3)
N_32F = normalizeImage(N_32F)
return N_32F
def _computeDetailNormal(self, N0_32F):
h, w = N0_32F.shape[:2]
W_32F = np.zeros((h, w))
# sigma_d = 2.0 * np.max(N0_32F[:, :, 2])
# W_32F = 1.0 - np.exp( - (N0_32F[:, :, 2] ** 2) / (sigma_d ** 2))
W_32F = 1.0 - N0_32F[:, :, 2]
W_32F *= 1.0 / np.max(W_32F)
W_32F = W_32F**1.5
A_c, b_c = amg_constraints.normalConstraints(W_32F, N0_32F)
A_L = amg_constraints.laplacianMatrix((h, w))
lambda_d = 2.0
A = A_c + lambda_d * A_L
b = b_c
N_32F = amg_solver.solve(A, b).reshape(h, w, 3)
N_32F = computeNz(N_32F.reshape(-1, 3)).reshape(h, w, 3)
N_32F = normalizeImage(N_32F)
return N_32F
def bilateralNormalSmoothing(image, normal):
sigma_xy = 1.0
xy = positionFeatures(image) / sigma_xy
Lab = LabFeatures(image)
foreground = foreGroundFeatures(image)
N = normal[:, :, :3].reshape(-1, 3)
LabxyN = np.concatenate((Lab, xy, N), axis=1)[foreground, :]
sigma_L = 1.0
LabxyN[:, 0] = LabxyN[:, 0] / sigma_L
LabxyN_sparse = shuffle(LabxyN, random_state=0)[:100]
N_smooth = np.array(N)
smooth = 10.0
f_x = np.vstack((LabxyN_sparse[:, :5].T, LabxyN_sparse[:, 5]))
f_y = np.vstack((LabxyN_sparse[:, :5].T, LabxyN_sparse[:, 6]))
f_z = np.vstack((LabxyN_sparse[:, :5].T, LabxyN_sparse[:, 7]))
Nx_rbf = Rbf(*(f_x), function='linear', smooth=smooth)
Ny_rbf = Rbf(*(f_y), function='linear', smooth=smooth)
Nz_rbf = Rbf(*(f_z), function='linear', smooth=smooth)
Labxy = LabxyN[:, :5]
#Lab_smooth[:, 0] = L_rbf(Labxy[:, 0], Labxy[:, 1], Labxy[:, 2], Labxy[:, 3], Labxy[:, 4])
N_smooth[foreground, 0] = Nx_rbf(*(Labxy.T))
N_smooth[foreground, 1] = Ny_rbf(*(Labxy.T))
N_smooth[foreground, 2] = Nz_rbf(*(Labxy.T))
h, w = image.shape[:2]
N_smooth = N_smooth.reshape((h, w, 3))
N_smooth = normalizeImage(N_smooth)
return N_smooth
def func(N):
N = normalizeImage(N, th)
return N
def func(N):
N = normalizeImage(N, th)
return N
def overviewFigure():
cmap_id = 10
colormap_file = colorMapFile(cmap_id)
num_rows = 1
num_cols = 5
w = 10
h = w * num_rows / num_cols
fig, axes = plt.subplots(figsize=(w, h))
font_size = 15
fig.subplots_adjust(left=0.02, right=0.98, top=0.96, bottom=0.02, hspace=0.05, wspace=0.05)
fig.suptitle("", fontsize=font_size)
plot_grid = SubplotGrid(num_rows, num_cols)
L = normalizeVector(np.array([-0.4, 0.6, 0.6]))
L_img = lightSphere(L)
shape_name = "ThreeBox"
Ng_data = shapeFile(shape_name)
Ng_data = loadNormal(Ng_data)
Ng_32F, A_8U = Ng_data
N0_file = shapeResultFile(result_name="InitialNormal", data_name=shape_name)
N0_data = loadNormal(N0_file)
N0_32F, A_8U = N0_data
M_32F = loadColorMap(colormap_file)
Cg_32F = ColorMapShader(M_32F).diffuseShading(L, Ng_32F)
borders=[0.6, 0.8, 0.92]
colors = [np.array([0.2, 0.2, 0.4]),
np.array([0.3, 0.3, 0.6]),
np.array([0.4, 0.4, 0.8]),
np.array([0.5, 0.5, 1.0])]
#Cg_32F = ToonShader(borders, colors).diffuseShading(L, Ng_32F)
#Cg_32F = cv2.GaussianBlur(Cg_32F, (0,0), 2.0)
sfs_method = ToonSFS(L, Cg_32F, A_8U)
sfs_method.setInitialNormal(N0_32F)
sfs_method.setNumIterations(iterations=40)
sfs_method.setWeights(w_lap=10.0)
sfs_method.run()
N_32F = sfs_method.normal()
I_32F = np.float32(np.clip(LdotN(L, N_32F), 0.0, 1.0))
I0_32F = np.float32(np.clip(LdotN(L, N0_32F), 0.0, 1.0))
C_32F = sfs_method.shading()
C0_32F = sfs_method.initialShading()
M_32F = sfs_method.colorMap().mapImage()
L1 = normalizeVector(np.array([0.0, 0.6, 0.6]))
L1_img = lightSphere(L1)
C1_32F = sfs_method.relighting(L1)
L2 = normalizeVector(np.array([0.5, 0.8, 0.6]))
L2_img = lightSphere(L2)
C2_32F = sfs_method.relighting(L2)
N_sil = silhouetteNormal(A_8U, sigma=7.0)
N_sil[:, :, 2] = N_sil[:, :, 2] ** 10.0
N_sil = normalizeImage(N_sil)
A_sil = 1.0 - N_sil[:, :, 2]
A_sil = to8U(A_sil)
N_xy = N_sil[:, :, 0] ** 2 + N_sil[:, :, 1] ** 2
A_sil[N_xy < 0.1] = 0
title = ""
plot_grid.showImage(setAlpha(Cg_32F, to32F(A_8U)), title)
plot_grid.showImage(normalToColor(N0_32F, A_8U), title)
plot_grid.showImage(setAlpha(C0_32F, to32F(A_8U)), title)
plot_grid.showImage(normalToColor(N_32F, A_8U), title)
plot_grid.showImage(setAlpha(C_32F, to32F(A_8U)), title)
# plot_grid.showImage(normalToColor(Ng_32F, A_8U), title)
#showMaximize()
file_path = shapeResultFile("Overview", "Overview")
fig.savefig(file_path, transparent=True)
file_path = shapeResultFile("Overview", "Cg")
saveRGBA(file_path, setAlpha(Cg_32F, to32F(A_8U)))
file_path = shapeResultFile("Overview", "L")
saveRGB(file_path, gray2rgb(to8U(L_img)))
file_path = shapeResultFile("Overview", "L1")
saveRGB(file_path, gray2rgb(to8U(L1_img)))
file_path = shapeResultFile("Overview", "L2")
saveRGB(file_path, gray2rgb(to8U(L2_img)))
file_path = shapeResultFile("Overview", "N0")
saveNormal(file_path, N0_32F, A_8U)
file_path = shapeResultFile("Overview", "N_sil")
saveNormal(file_path, N_sil, A_sil)
file_path = shapeResultFile("Overview", "N")
saveNormal(file_path, N_32F, A_8U)
file_path = shapeResultFile("Overview", "C0")
saveRGBA(file_path, setAlpha(C0_32F, to32F(A_8U)))
file_path = shapeResultFile("Overview", "C")
saveRGBA(file_path, setAlpha(C_32F, to32F(A_8U)))
file_path = shapeResultFile("Overview", "C1")
saveRGBA(file_path, setAlpha(C1_32F, to32F(A_8U)))
file_path = shapeResultFile("Overview", "C2")
saveRGBA(file_path, setAlpha(C2_32F, to32F(A_8U)))
file_path = shapeResultFile("Overview", "I")
saveRGBA(file_path, setAlpha(gray2rgb(I_32F), to32F(A_8U)))
file_path = shapeResultFile("Overview", "I0")
saveRGBA(file_path, setAlpha(gray2rgb(I0_32F), to32F(A_8U)))
file_path = shapeResultFile("Overview", "M")
saveRGB(file_path, M_32F)
def computeInitialNormal(self):
if self._N_lumo.shape != self._N_b_smooth.shape:
return
self._N0_b_32F = normalizeImage(bumpMapping(np.array(self._N_lumo), self._N_b_smooth))
self._N0_d_32F = normalizeImage(bumpMapping(np.array(self._N_lumo), self._N_d_smooth))
