def forward(self, prefix = "/tmp/", N=5, T=2):
'''A Motion Compensated Discrete Wavelet Transform.
Compute the MC 1D-DWT. The input video (as a sequence of
1-levels decompositions) must be stored in disk in the
directory <prefix>, and the output (as a 1-levels MC
decompositions) will generated in the same directory.
Input
-----
prefix : str
Localization of the input/output images. Example:
"/tmp/".
N : int
Number of decompositions to process.
T : int
Number of levels of the MCDWT (temporal scales). Controls
the GOP size.
T | GOP_size
----+-----------
0 | 1
1 | 2
2 | 4
3 | 8
4 | 16
5 | 32
: | :
P : int
Predictor to use:
1 --> Average Predictor.
2 --> Weighted Average Predictor.
Returns
-------
None.
'''
x = 2
for t in range(T): # Temporal scale
i = 0
aL, aH = decomposition.read(prefix, "{:03d}".format(0))
while i < (N//x):
bL, bH = decomposition.read(prefix, "{:03d}".format(x*i+x//2))
cL, cH = decomposition.read(prefix, "{:03d}".format(x*i+x))
bH = self.__forward_butterfly(aL, aH, bL, bH, cL, cH)
decomposition.writeH(bH, prefix, "{:03d}".format(x*i+x//2))
aL, aH = cL, cH
i += 1
x *= 2
def forward(self, prefix="/tmp/", N=5, T=2):
'''Forward MCOLP.
Compute a MC 1D-DWT, estimating in the L subbands and
compensaing in the H subbands of the Orthogonal Laplacian
Pyramid. The input video (as a sequence of 1-iteration
decompositions) must be stored in disk in the directory
<prefix>, and the output (as a 1-iteration MC decompositions)
will generated in the same directory.
Input
-----
prefix : str
Localization of the input/output images. Example:
"/tmp/".
N : int
Number of decompositions to process.
T : int
Number of iterations of the MCOLP (temporal scales).
Controls the GOP size.
T | GOP_size
----+----------
0 | 1
1 | 2
2 | 4
3 | 8
4 | 16
5 | 32
: | :
Returns
-------
The output motion compensated decompositions.
'''
x = 2
for t in range(T): # Temporal scale
#print("a={}".format(0), end=' ')
i = 0
aL, aH = decomposition.read(prefix, "{:03d}".format(0))
while i < (N // x):
#print("b={} c={}".format(x*i+x//2, x*i+x))
bL, bH = decomposition.read(prefix,
"{:03d}".format(x * i + x // 2))
cL, cH = decomposition.read(prefix, "{:03d}".format(x * i + x))
residue_bH = self.__forward_butterfly(aL, aH, bL, bH, cL, cH)
decomposition.writeH(residue_bH, prefix,
"{:03d}".format(x * i + x // 2))
aL, aH = cL, cH
#print("a={}".format(x*i+x), end=' ')
i += 1
x *= 2
def forward(self, prefix="/tmp/", N=5, I=2):
'''Forward MCDWT.
Compute the MC 1D-DWT. The input video (as a sequence of
1-iteration decompositions) must be stored in disk in the
directory <prefix>, and the output (as a 1-iteration MC
decompositions) will generated in the same directory.
Input
-----
prefix : str
Localization of the input/output images. Example:
"/tmp/".
N : int
Number of decompositions to process.
I : int
Number of iterations of the MCDWT (temporal scales).
Controls the GOP size.
I | GOP_size
----+----------
0 | 1
1 | 2
2 | 4
3 | 8
4 | 16
5 | 32
: | :
Returns
-------
The output motion compensated decompositions.
'''
x = 2
for t in range(I): # Temporal scale
i = 0
aL, aH = decomposition.read(prefix, "{:03d}".format(0))
while i < (N // x):
bL, bH = decomposition.read(prefix,
"{:03d}".format(x * i + x // 2))
cL, cH = decomposition.read(prefix, "{:03d}".format(x * i + x))
residue_bH = self.__forward_butterfly(aL, aH, bL, bH, cL, cH)
decomposition.writeH(residue_bH, prefix,
"{:03d}".format(x * i + x // 2))
aL, aH = cL, cH
i += 1
x *= 2
def backward(self, prefix = "/tmp/", N=5, T=2):
x = 2**T
for t in range(T): # Temporal scale
i = 0
aL, aH = decomposition.read(prefix, "{:03d}".format(0))
while i < (N//x):
bL, bH = decomposition.read(prefix, "{:03d}".format(x*i+x//2))
cL, cH = decomposition.read(prefix, "{:03d}".format(x*i+x))
bH = self.__backward_butterfly(aL, aH, bL, bH, cL, cH)
decomposition.writeH(bH, "{:03d}".format(x*i+x//2))
aL, aH = cL, cH
i += 1
x //=2
def backward(self, prefix="/tmp/", N=5, T=2):
'''Backward MCDWT.
Compute the inverse MC 1D-DWT. The input sequence of
1-iteration MC decompositions must be stored in disk in the
directory <prefix>, and the output (as a 1-iteration
decompositions) will generated in the same directory.
Input
-----
prefix : str
Localization of the input/output images. Example:
"/tmp/".
N : int
Number of decompositions to process.
T : int
Number of iterations of the MCDWT (temporal scales).
Returns
-------
The sequence of 1-iteration decompositions.
'''
x = 2**T
for t in range(T): # Temporal scale
i = 0
aL, aH = decomposition.read(prefix, "{:03d}".format(0))
while i < (N // x):
bL, residue_bH = decomposition.read(
prefix, "{:03d}".format(x * i + x // 2))
cL, cH = decomposition.read(prefix, "{:03d}".format(x * i + x))
bH = self.__backward_butterfly(aL, aH, bL, residue_bH, cL, cH)
decomposition.writeH(bH, prefix,
"{:03d}".format(x * i + x // 2))
aL, aH = cL, cH
i += 1
x //= 2
'''MODIFICA LA SUBBANDA HH PONIENDOLA EN NEGRO CON UN PUNTO BLANCO EN EL CENTRO'''
for i in range(args.N):
LH, HL, HH = decomposition.readH("{}{:03d}".format(args.decompositions, i))
LL = decomposition.readL("{}{:03d}".format(args.decompositions, i))
y = math.ceil(HH.shape[0]/2)
x = math.ceil(HH.shape[1]/2)
HH = HH * 0
HH[x][y][0] = 255
HH[x][y][1] = 255
HH[x][y][2] = 255
decomposition.writeH([LH,HL,HH],"{}{:03d}".format(args.decompositions, i))
'''SE RECONSTRUYE LA IMAGEN Y SE VUELVE A DESCOMPONERLA'''
d.backward(args.decompositions, '/tmp/recons_MDWT_', args.N)
d.forward('/tmp/recons_MDWT_', args.decompositions, args.N)
for i in range(args.N):
print("ESPACIAL NIVEL {}\n"
"********\n".format(i))
LH, HL, HH = decomposition.readH("{}{:03d}".format(args.decompositions, i))
LL = decomposition.readL("{}{:03d}".format(args.decompositions, i))
path += '/LL/'
