Implements a shift register. Input is scalar, out is a vector whose first coordinate is the most recent input

The input is FIR filtered, the subsampled by retaining only one value out of two.

Output rate is half of the input's.


Pure discrete delay..


A symmetric dead zone centered on zero. Can be used in wavelet denoising.


Subsampled wavelet analysis. The input is processed by two filters and the outputs are subsampled using the block on the left.


Shift invariant wavelet analysis. The input is simply processed by two FIR filters which are dilated when the scale increases.


The details are suitably scaled and a finite difference is performed on the LoRes signal. This is used to compute the derivative of a signal using its wavelet coefficients.

8 scales are assumed to exist, e.g., Details has a width of 8.

see pdf document for mathematical background.


Oversamples a signal by inserting a zero between every successive pair of samples.

Ouput rate is double of the input's.


The signal is fist oversampled by zero insertion (see left block) and filtered.


The delay block is used to synchronize a signal with another one which is filtered by a non causal filter and at at another scale.


Input is set to zero if its absolute value is below a given threshold. Can be used in wavelet denoising.


Signal reconstruction from the analysis above. The two inputs are "punched and filtered" and added to produce the output.


Shift invariant wavelet synthesis. The input is simply processed by two dilated FIR filters and their sum is devided by 2 two produce the output.


Finite differente operator. It is a special implementattion of a FIR filter