def forward(X,h = [0.0625,0.25,0.375,0.25,0.0625],J = 1):
nX = np.shape(X)
Lh = np.size(h)
W = sp2.Starlet2D(nX[1],nX[2],nX[0],J,Lh).forward_omp(np.real(X),np.array(h))
return W
def backward(W,h = [0.0625,0.25,0.375,0.25,0.0625]):
nX = np.shape(W)
Lh = np.size(h)
rec = sp2.Starlet2D(nX[1],nX[2],nX[0],nX[3]-1,Lh).backward_omp(np.real(W))
return rec
def adjoint1d(W,h = [0.0625,0.25,0.375,0.25,0.0625]):
nX = np.shape(W)
Lh = np.size(h)
W = sp2.Starlet2D(nX[1],1,nX[0],nX[2]-1,Lh).adjoint1d(np.real(W),np.array(h))
return W
def backward(data, consts=[0.0625, 0.25, 0.375, 0.25, 0.0625]):
""" multichannel starlet recomposition """
nsz = np.shape(data)
csz = np.size(consts)
rec = sp2.Starlet2D(nsz[1], nsz[2], nsz[0], nsz[3] - 1,
csz).backward_omp(np.real(data))
return rec
def forward(data, consts=[0.0625, 0.25, 0.375, 0.25, 0.0625], jscale=1):
""" multichannel starlet decomposition """
nsz = np.shape(data)
csz = np.size(consts)
trans = sp2.Starlet2D(nsz[1], nsz[2], nsz[0], jscale, csz).forward_omp(np.real(data), \
np.array(consts))
return trans
def forward1d(X,h = [0.0625,0.25,0.375,0.25,0.0625],J = 1):
"""
soit x le vecteur qu'on veut transformer en starlet
X=np.array(x).reshape((1, np.shape(x)))
W : de taille 1*(taille de x)*J+1
"""
nX = np.shape(X)
Lh = np.size(h)
W = sp2.Starlet2D(nX[1],1,nX[0],J,Lh).forward1d_omp(np.real(X),np.array(h))
return W
def forward1d(X,h = [0.0625,0.25,0.375,0.25,0.0625],J = 1):
nX = np.shape(X)
Lh = np.size(h)
W = sp2.Starlet2D(nX[1],1,nX[0],J,Lh).forward1d_omp(np.real(X),np.array(h))
normFact = np.array([ 0.72358037, 0.28547571, 0.17796025, 0.12221233, 0.08578265,
0.06062301, 0.04293804, 0.0302886 , 0.02141538, 0.01511499,
0.01055062, 0.00739429, 0.00522072, 0.00372874, 0.00252544,
0.00280893])
for ii in range(np.shape(W)[2]-1): # Warning: the coarse scale is not normalized, tested
W[:,:,ii] = W[:,:,ii]/normFact[ii]
return W
