def __init__(
self,
huefile,
saturationfile,
valuefile,
huemin,
saturationmin,
valuemin,
huemax,
saturationmax,
valuemax,
gamma=1.0,
bb=[0, 0, -1, -1],
):
self.hdata = FitsData(huefile, bb)
self.sdata = FitsData(saturationfile, bb)
self.vdata = FitsData(valuefile, bb)
# use the original data to calculate the RGB values
hue = (numarray.ravel(self.hue.hdata) - huemin) / (huemax - huemin)
hue = (numarray.ravel(self.hue.hdata) - huemin) / (huemax - huemin)
hue = (numarray.ravel(self.hue.hdata) - huemin) / (huemax - huemin)
self.red = MyImage.__init__(self, "L", self.hdata.size)
self.green = MyImage.__init__(self, "L", self.hdata.size)
self.blue = MyImage.__init__(self, "L", self.hdata.size)
MyImage.__init__(self, "RGB", self.hdata.size)
self.im.putband(self.hue.im, 0)
self.im.putband(self.saturation.im, 1)
self.im.putband(self.value.im, 2)
def get_optic_flow(im0,im1,pix,wnd,dt):
"""
Compute optic flow at pixel location
Inputs:
im0 -- image 0
im1 -- image 1
pix -- pixel locations
wnd -- computation window
dt -- time step between images
"""
# Get index location of pixel
i0,i1 = pix[1], pix[0]
# Get regions for optic flow calculation
E1_w_bndry = im1[i0-wnd-2:i0+wnd+3,i1-wnd-2:i1+wnd+3]
E0 = im0[i0-wnd:i0+wnd+1,i1-wnd:i1+wnd+1]
# Cast to Float for computation
E1_w_bndry = nx.asarray(E1_w_bndry, nx.Float32)
E0 = nx.asarray(E0, nx.Float32)
E1 = E1_w_bndry[2:-2,2:-2]
# Gaussian Filter
#std = 1.5
#E1_w_bndry = gaussian_filter(E1_w_bndry,std)
#E1 = E1_w_bndry[2:-2,2:-2]
#E0 = gaussian_filter(E0, std)
# Get time derivatives
dE_dt = (E1-E0)/dt
# Get spacial derivatives
Ex_w_bndry, Ey_w_bndry = lsq_deriv(E1_w_bndry)
Ex = Ex_w_bndry[2:-2,2:-2]
Ey = Ey_w_bndry[2:-2,2:-2]
# Create Problem matrices
n = 2*wnd+1
A = nx.concatenate((nx.ravel(Ex)[:,None], nx.ravel(Ey)[:,None]), axis=1)
b = -nx.ravel(dE_dt)
## Solution method 1 (fastest?) ------------------------
#At = nx.transpose(A)
#AtA = nx.matrixmultiply(At,A)
#Atb = nx.matrixmultiply(At,b)
#try:
# x = solve(AtA,Atb)
#except:
# x = nx.array([0.0,0.0])
# Solution method 2 ------------------------------------
# New method
x,resids,rank,s = scipy.linalg.lstsq(A,b)
return x[0], x[1]
def get_optic_flow(im0, im1, pix, wnd, dt):
"""
Compute optic flow at pixel location
Inputs:
im0 -- image 0
im1 -- image 1
pix -- pixel locations
wnd -- computation window
dt -- time step between images
"""
# Get index location of pixel
i0, i1 = pix[1], pix[0]
# Get regions for optic flow calculation
E1_w_bndry = im1[i0 - wnd - 2:i0 + wnd + 3, i1 - wnd - 2:i1 + wnd + 3]
E0 = im0[i0 - wnd:i0 + wnd + 1, i1 - wnd:i1 + wnd + 1]
# Cast to Float for computation
E1_w_bndry = nx.asarray(E1_w_bndry, nx.Float32)
E0 = nx.asarray(E0, nx.Float32)
E1 = E1_w_bndry[2:-2, 2:-2]
# Gaussian Filter
#std = 1.5
#E1_w_bndry = gaussian_filter(E1_w_bndry,std)
#E1 = E1_w_bndry[2:-2,2:-2]
#E0 = gaussian_filter(E0, std)
# Get time derivatives
dE_dt = (E1 - E0) / dt
# Get spacial derivatives
Ex_w_bndry, Ey_w_bndry = lsq_deriv(E1_w_bndry)
Ex = Ex_w_bndry[2:-2, 2:-2]
Ey = Ey_w_bndry[2:-2, 2:-2]
# Create Problem matrices
n = 2 * wnd + 1
A = nx.concatenate((nx.ravel(Ex)[:, None], nx.ravel(Ey)[:, None]), axis=1)
b = -nx.ravel(dE_dt)
## Solution method 1 (fastest?) ------------------------
#At = nx.transpose(A)
#AtA = nx.matrixmultiply(At,A)
#Atb = nx.matrixmultiply(At,b)
#try:
# x = solve(AtA,Atb)
#except:
# x = nx.array([0.0,0.0])
# Solution method 2 ------------------------------------
# New method
x, resids, rank, s = scipy.linalg.lstsq(A, b)
return x[0], x[1]
def __init__(self, file1, file2, fmin, fmax, gamma=1.0, bb=[0, 0, -1, -1], minquality=1.0):
self.numerator = FitsData(file1, bb)
self.denominator = FitsData(file2, bb)
self.size = self.numerator.size
if self.size != self.denominator.size:
raise IndexError, "Incompatible sizes in RatioImage!"
MyImage.__init__(self, "L", self.size)
if self.ratiofunc == None:
self.fitsdata = self.numerator.fitsdata / self.denominator.fitsdata
else:
self.fitsdata = self.ratiofunc(self.numerator.fitsdata, self.denominator.fitsdata)
self.flatnormaldata = (numarray.ravel(self.fitsdata) - fmin) / (fmax - fmin)
self.flatnormaldata = numarray.clip(self.flatnormaldata, 0.0, 1.0)
if gamma != 1.0:
self.flatnormaldata = self.flatnormaldata ** (1.0 / gamma)
self.putdata(self.flatnormaldata, scale=255.0)
self.quality = numarray.where(self.denominator.fitsdata > minquality, 1, 0)
self.quality = numarray.where(self.numerator.fitsdata > minquality, self.quality, 0)
self.alpha = Image.new("L", self.size)
self.alpha.putdata(numarray.ravel(self.quality), scale=255)
self.im = Image.composite(self, Image.new("L", self.size, 255), self.alpha).im
def __init__(self, fitsfile, fmin, fmax, gamma=1.0, bb=[0, 0, -1, -1], icut=None, cutaxis=1, takelog=0):
FitsData.__init__(self, fitsfile, bb, icut, cutaxis)
MyImage.__init__(self, "L", self.size)
# 06 Dec 2010: add new attributes: min, max to allow outside inspection
# 05 Dec 2010: If fmin or fmax is None, then the respective limit is found from the data
self.min = fmin if not fmin is None else self.fitsdata.min()
self.max = fmax if not fmax is None else self.fitsdata.max()
if takelog:
self.flatnormaldata = (numarray.ravel(numarray.log10(self.fitsdata)) - numarray.log10(self.min)) / (
numarray.log10(self.max) - numarray.log10(self.min)
)
else:
self.flatnormaldata = (numarray.ravel(self.fitsdata) - self.min) / (self.max - self.min)
# 09 Sep 2005: How on earth did this ever work before? It is necessary
# to clip fitsdata to [fmin,fmax] before applying gamma correction, or
# else we get sqrt of negative numbers....
# Aha, there is a smart way to do this!
self.flatnormaldata = numarray.clip(self.flatnormaldata, 0.0, 1.0)
if gamma != 1.0:
self.flatnormaldata = self.flatnormaldata ** (1.0 / gamma)
self.putdata(self.flatnormaldata, scale=255.0)
def numarray_to_cvx(*args):
"""Returns CVX matrices from NumArray matrices
Note that NumArray one-dimensional arrays are transformed as row vectors
in CVX. See/Run test_numarray_cvx_conversions() for details.
"""
matrices = []
for a in args:
a = numarray_asmatrix(a)
m, n = a.shape
a = numarray.ravel( numarray.transpose(a) )
matrices.append( cvx.matrix(a, (m,n), 'd') )
if len(matrices)==1:
return matrices[0]
return matrices
def numarray_to_cvx(*args):
"""Returns CVX matrices from NumArray matrices
Note that NumArray one-dimensional arrays are transformed as row vectors
in CVX. See/Run test_numarray_cvx_conversions() for details.
"""
matrices = []
for a in args:
a = numarray_asmatrix(a)
m, n = a.shape
a = numarray.ravel(numarray.transpose(a))
matrices.append(cvx.matrix(a, (m, n), 'd'))
if len(matrices) == 1:
return matrices[0]
return matrices
