def recon_2d_haar_folder_size(data,row_tree,col_tree,threshold=0.0):
"""
Reconstruction of a function in two variables (data) by the following:
Find the bi-Haar expansion of data in terms of the row_tree/col_tree
basis. Then set all coefficients corresponding to folders of
size < threshold to 0.0, and perform the inverse transform.
"""
coefs, folder_sizes = haar.bihaar_transform(data,row_tree,col_tree,True)
coefs[folder_sizes < threshold] = 0.0
return haar.inverse_bihaar_transform(coefs,row_tree,col_tree)
def recon_2d_haar_folder_size(data, row_tree, col_tree, threshold=0.0):
"""
Reconstruction of a function in two variables (data) by the following:
Find the bi-Haar expansion of data in terms of the row_tree/col_tree
basis. Then set all coefficients corresponding to folders of
size < threshold to 0.0, and perform the inverse transform.
"""
coefs, folder_sizes = haar.bihaar_transform(data, row_tree, col_tree, True)
coefs[folder_sizes < threshold] = 0.0
return haar.inverse_bihaar_transform(coefs, row_tree, col_tree)
def recon_2d_sure(data,row_tree,col_tree,estimated_var=1.0):
"""
Reconstruction of a function in two dimensions using SURE wavelet shrinkage.
"""
bihaar_coefs = haar.bihaar_transform(data, row_tree, col_tree)
new_bihaar_coefs = np.zeros(bihaar_coefs.shape)
row_levels = haar.level_correspondence(row_tree)
col_levels = haar.level_correspondence(col_tree)
coef_levels = np.add.outer(row_levels,col_levels)
for level in np.unique(coef_levels):
hc = bihaar_coefs[coef_levels==level]
if len(hc) == 1:
t = 0
else:
x = np.arange(0,6.0,0.1)
estimates = []
for threshold in x:
estimate = sure(hc,threshold,estimated_var)
estimates.append(estimate[0])
t = x[np.argmin(estimates)]
new_bihaar_coefs[coef_levels==level] = shrink_coefs(hc,t)
return haar.inverse_bihaar_transform(new_bihaar_coefs, row_tree, col_tree)
def recon_2d_sure(data, row_tree, col_tree, estimated_var=1.0):
"""
Reconstruction of a function in two dimensions using SURE wavelet shrinkage.
"""
bihaar_coefs = haar.bihaar_transform(data, row_tree, col_tree)
new_bihaar_coefs = np.zeros(bihaar_coefs.shape)
row_levels = haar.level_correspondence(row_tree)
col_levels = haar.level_correspondence(col_tree)
coef_levels = np.add.outer(row_levels, col_levels)
for level in np.unique(coef_levels):
hc = bihaar_coefs[coef_levels == level]
if len(hc) == 1:
t = 0
else:
x = np.arange(0, 6.0, 0.1)
estimates = []
for threshold in x:
estimate = sure(hc, threshold, estimated_var)
estimates.append(estimate[0])
t = x[np.argmin(estimates)]
new_bihaar_coefs[coef_levels == level] = shrink_coefs(hc, t)
return haar.inverse_bihaar_transform(new_bihaar_coefs, row_tree, col_tree)
