Crate ndwt 

Wavelet transforms for real and complex signals.

This crate provides two families of wavelet transform:

• DWT (dwt) — the classic Discrete Wavelet Transform via convolution/subsampling.
• LWT (lwt) — the Lifting Wavelet Transform, an in-place factorisation of the DWT.

Both families support 1-D and N-D transforms, multi-level decomposition, periodic and general boundary conditions, SIMD acceleration (via pulp), and optional rayon parallelism (feature rayon).

Also notable is that we provide adjoint operations for all of the forward and inverse transforms that respect the boundary extension modes, enabling these transforms to be cleanly used in optimization problems.

Wavelet families

All concrete wavelet types live in the sub-modules below. Each type is a zero-size marker struct that carries the filter coefficients as compile-time constants.

Module	Wavelets
daubechies	Daubechies 1–10
symlet	Symlets 4–6
coiflet	Coiflets 1–3
bior	Biorthogonal & CDF variants

The Wavelet enum lets you select a wavelet at runtime without generics.

Quick start

use ndwt::{Wavelet, boundarys::BoundaryCondition, dwt::driver::WaveletTransform};

// Build a DWT driver for Daubechies-4 with zero boundary padding.
let xfm: = WaveletTransform::new(Wavelet::Daubechies4, BoundaryCondition::Zero);

let input = vec![1.0_f64; 128];
let nsd = ndwt::dwt::get_outlen(8, input.len());
let mut approx = vec![0.0; nsd];
let mut detail = vec![0.0; nsd];
xfm.forward_1d(&input, &mut approx, &mut detail);

Modules

bior

Biorthogonal and CDF wavelet families.

boundarys

Boundary extension strategies for wavelet transforms.

coiflet

Coiflet wavelets (orders 1–3).

daubechies

Daubechies orthogonal wavelets (orders 1–10).

dwt

Discrete Wavelet Transform (DWT) via direct convolution and subsampling.

iter

N-dimensional lane iteration over flat slices and ndarray arrays.

lwt

Lifting Wavelet Transform (LWT).

simd

SIMD extension traits and CPU-dispatch infrastructure used by the wavelet kernels.

symlet

Symlet near-symmetric wavelets (orders 4–6).

utils

Array layout utilities for wavelet sub-band manipulation.

Enums

Wavelet

Wavelet family selector.

Traits

ChunkWidth

Marker trait setting N to be a good default value for the driver chunk sizes for type T.

MulScalarAdd

Fused multiply-add: self * a + b.

Transformable

Element type that can participate in wavelet filter convolutions.

Functions

max_level

Compute the maximum number of decomposition levels for a filter of width N applied to a signal of length n.

max_level_nd

Compute the maximum number of decomposition levels for a filter of width N applied to an n-dimensional signal with shape along axes.