def distance_vs_roughness(n_init, gamma, ba):
'''
Computing the distance between the target and the final curve with respect
to roughness of circular wave. In this function we fix the number of
initial points, gamma, beta/alpha and the peiodicity
'''
lwr_lst = np.linspace(10, 50, 10)
spl = spline_initialize(R=200, n_init=n_init)
dist = np.zeros_like(lwr_lst)
plt.rcParams["font.family"] = "serif"
for i, data in enumerate(lwr_lst):
spl_c = copy.deepcopy(spl)
pts_ref = get_pts_from_sgeom_vague(R=50, lwr=data)
img, _ = create_vague(R=50, lwr=data)
opti = optimize(spline=spl_c,
method='slsqp',
img=img,
jac=None,
ba_ratio=data,
gamma=gamma)
lst = np.reshape(opti.x, (len(opti.x) / 2, 2))
spl.set_ctrl_pts(lst, only_free=True)
pts = spl(np.linspace(0, 1, 10000))
dist[i] = distance_between_curves(pts_ref, pts)
plt.plot(lwr_lst, dist)
plt.title(u"Nombre de points de contrôle = " + str(n_init) +
r', $\gamma = $' + str(gamma) + r', $\beta/\alpha = $' + str(ba))
plt.xlabel(u'Rugosité')
plt.ylabel('Distance')
plt.show()
def distance_vs_gamma(ba, n_init):
'''
Computing the distance between the target and the final curve with respect
to gamma. In this function we fix the beta/alpha and number of initial
points
'''
spl = spline_initialize(R=200, n_init=n_init)
img, imsh = create_vague(R=50)
gamma_lst = np.linspace(40, 200, 20)
dist = np.zeros_like(gamma_lst)
pts_ref = get_pts_from_sgeom_vague(R=50)
plt.rcParams["font.family"] = "serif"
for i, dat in enumerate(gamma_lst):
spl_c = copy.deepcopy(spl)
opti = optimize(spline=spl_c,
method='slsqp',
img=img,
gamma=dat,
ba_ratio=ba,
jac=None)
lst = np.reshape(opti.x, (len(opti.x) / 2, 2))
spl_c.set_ctrl_pts(lst, only_free=True)
pts = spl_c(np.linspace(0, 1, 10000))
dist[i] = distance_between_curves(pts_ref, pts)
plt.plot(gamma_lst, dist)
plt.title(r"$\beta/\alpha = $" + str(ba) +
u', nombre de points de contrôle = ' + str(n_init))
plt.xlabel(r'$\gamma$')
plt.ylabel('Distance')
plt.show()
def distance_vs_n_init(ba, gamma):
'''
Computing the distance between the target and the final curve with respect
to number of initial points. In this function we fix the beta/alpha and
gamma
'''
nbr_lst = np.linspace(5, 40, 20)
img, _ = create_vague(R=50)
dist = np.zeros_like(nbr_lst)
nbr_lst = nbr_lst.astype(int)
pts_ref = get_pts_from_sgeom_vague(R=50)
plt.rcParams["font.family"] = "serif"
for i, data in enumerate(nbr_lst):
spl = spline_initialize(R=200, n_init=data)
opti = optimize(spline=spl,
method='slsqp',
img=img,
gamma=gamma,
ba_ratio=ba,
jac=None)
lst = np.reshape(opti.x, (len(opti.x) / 2, 2))
spl.set_ctrl_pts(lst, only_free=True)
pts = spl(np.linspace(0, 1, 10000))
dist[i] = distance_between_curves(pts_ref, pts)
plt.plot(nbr_lst, dist)
plt.title(r"$\beta/\alpha = $" + str(ba) + r', $\gamma = $' + str(gamma))
plt.xlabel(u'Nombre de points de contrôle')
plt.ylabel('Distance')
plt.show()
def optimize_image(jaco):
'''
A module of optimisation for the Active Contour. However, it is only
effective for the traditional method of jacobian
'''
spl = spline_initialize(R=200, n_init=7)
img, imsh = create_vague(50)
imsh.plot(label='Vrai contour')
spl.plot_spline(label='Courbe initiale')
st = timeit.default_timer()
opti = optimize(spline=spl,
method='slsqp',
img=img,
gamma=75.,
ba_ratio=100,
jac=jaco)
# ed = timeit.default_timer()
# print ed - st
lst = np.reshape(opti.x, (len(opti.x) / 2, 2))
spl.set_ctrl_pts(lst, only_free=True)
spl.plot_spline(label='Courbe finale')
[x, y] = lst.T
plt.scatter(x, y)
plt.title(u"Nombre de points de contrôle = 7" + r', $\gamma = 75$' +
r', $\beta/\alpha = 100$')
# plt.title(u'Rugosité = 60')
plt.legend()
plt.show()
def distance(ba, gamma, n_init, per=np.pi / 3):
'''
Computing the distance between the final curve and the contour with respect
to parameters like ba_ratio, gamma, number of initial points
'''
ref_pts = get_pts_from_sgeom_circle(R=50)
spl = spline_initialize(R=200, n_init=n_init)
img, _ = create_circle(R=50)
opti = optimize(spline=spl,
method='slsqp',
img=img,
jac=None,
ba_ratio=ba,
gamma=gamma)
lst = np.reshape(opti.x, (len(opti.x) / 2, 2))
spl.set_ctrl_pts(lst, only_free=True)
pts = spl(np.linspace(0, 1, 3000))
dist = distance_between_curves(ref_pts, pts)
print dist