Expands dimensions of 2D matrices to match the required tensor dimensions in a TF graph, and wrapping them as TF constants. Args:

Returns filter coefficients at requested indeces. Indeces are offset by the filter origin Args: item (int or sli

Create a trainable filter initialized with given filter coefficients Args: initial_coeffs (np.ndarray): Initial filte

Create a new wavelet based on specified filters Args: decomp_lp (Filter): Filter coefficients for decomposition low p

Create a new trainable wavelet initialized as specified wavelet Args: wavelet (Wavelet): Starting point for the

Cyclic convolution Args: input_node: Input signal (3-tensor [batch, width, in_channels]) filter_: Filter Returns:

Computes the DWT of a 1D signal. Args: signal (np.ndarray): A 1D array to compute DWT of. wavelet (str): Type of

Constructs a TF computational graph computing the 2D DWT of an input signal. Args: input_node: A 3D tensor containing the signal

Computes the DWT of a 2D signal. Args: signal (np.ndarray): A 2D array to compute DWT of. wavelet (str): Type of

Get a wavelet based on the wavelets name. Args: wavelet_name (str): Name of the wavelet ('haar', 'db1', 'db2', 'db3' or 'db4').

Computes the inverse DWT of a 1D signal. Args: signal (np.ndarray): A 1D array to compute IDWT of. wavelet (str):

Constructs a TF graph that computes the 2D inverse DWT for a given wavelet. Args: input_node (tf.placeholder): Input sig

Computes the inverse DWT of a 2D signal. Args: signal (np.ndarray): A 2D array to compute iDWT of. wavelet (str):