def testImageDctIsOrthonormal(self):
"""Test that <im0, im1> = <image_dct(im0), image_dct(im1)>."""
for _ in range(4):
im0 = np.float32(np.random.uniform(size=(4, 4, 2)))
im1 = np.float32(np.random.uniform(size=(4, 4, 2)))
dct_im0 = util.image_dct(im0)
dct_im1 = util.image_dct(im1)
prod1 = tf.reduce_sum(im0 * im1)
prod2 = tf.reduce_sum(dct_im0 * dct_im1)
self.assertAllClose(prod1, prod2)
def transform_to_mat(self, x):
"""Transforms a batch of images to a matrix."""
assert len(x.shape) == 4
x = torch.as_tensor(x)
if self.color_space == 'YUV':
x = util.rgb_to_syuv(x)
# If `color_space` == 'RGB', do nothing.
# Reshape `x` from
# (num_batches, width, height, num_channels) to
# (num_batches * num_channels, width, height)
_, width, height, num_channels = x.shape
x_stack = torch.reshape(x.permute(0, 3, 1, 2), (-1, width, height))
# Turn each channel in `x_stack` into the spatial representation specified
# by `representation`.
if self.representation in wavelet.generate_filters():
x_stack = wavelet.flatten(
wavelet.rescale(
wavelet.construct(x_stack, self.wavelet_num_levels,
self.representation),
self.wavelet_scale_base))
elif self.representation == 'DCT':
x_stack = util.image_dct(x_stack)
# If `representation` == 'PIXEL', do nothing.
# Reshape `x_stack` from
# (num_batches * num_channels, width, height) to
# (num_batches, num_channels * width * height)
x_mat = torch.reshape(
torch.reshape(x_stack, (-1, num_channels, width, height)).permute(
0, 2, 3, 1), [-1, width * height * num_channels])
return x_mat
def testImageDctIsCorrect(self):
"""Tests that image_dct() is volume preserving."""
im = np.float32(np.random.uniform(size=(3, 8, 8)))
im_dct = util.image_dct(im)
dct_y = tf.transpose(tf.spectral.dct(im, type=2, norm='ortho'),
[0, 2, 1])
dct_x = tf.transpose(tf.spectral.dct(dct_y, type=2, norm='ortho'),
[0, 2, 1])
im_dct_true = dct_x
self.assertAllClose(im_dct, im_dct_true, atol=1e-5, rtol=1e-5)
def testImageDctRoundTrip(self):
"""Tests that image_idct(image_dct(x)) == x."""
image = np.float32(np.random.uniform(size=(32, 32, 3)))
image_recon = util.image_idct(util.image_dct(image))
self.assertAllClose(image, image_recon)