A Discrete Fourier Transform (DFT), a Fast Wavelet Transform (FWT), and a Wavelet Packet Transform (WPT) algorithm in 1-D, 2-D, and 3-D using normalized orthogonal (orthonormal) Haar, Coiflet, Daubechie, Legendre and normalized biorthognal wavelets in Java.
Java implementation of a Discrete Fourier Transform (DFT), a Fast Wavelet Transform (FWT), and a Wavelet Packet Transform (WPT) algorithm. All algorithms are available in 1-D, 2-D, and 3-D. The wavelet transform algorithms are using normalized orthogonal or if available orthonormal wavelets. The comon wavelets like Haar, Coiflet, Daubechies, Symlets, and Legendre are available. Additionally there are also some Bi-Orthogonal and unusal wavelets implemented - in total around 50 wavelets.
The implementation of JWave is based on several software design patterns and - hopefully - appears therefore user-friendly.
Have a look at the little how to down the page or try looking at the unit tests.
However, the bare algorithms of JWave do only support data sampled by 2^p | p E N; e.g. 2, 4, 8, 16, .. 128, 256, 512, 1024, .. and so on. Please use the AncientEgyptianDecomposition class for odd samplings (most frequently asked question)! You can find it down the page.
Why do the results (hilbert spaces) look totally different to those from my matlab or some other implementation I found on net?!
In most cases, other libraries construct the orthogonal / orthonormal bases of the wavelet and scaling function in a different way. Especially for those bases of multiple dimension or wavelets of higher dimension, repectively. But does this hurt?
Totally not!
The why can be found in mathematics. Due to using some orthogonal transform (or better, an orthogonal base), it is up to oneself how to construct this base (as long it stays orthogonal over all dimensions). Next it is also up to oneself how to apply the sequence of the transform steps. Both does not influence any performance of the wavelet transforms! But again why?
The base stays orthogonal, and one's data is unit rotated and mirrowed differently, which makes a long story short.
Additionally the application of the transform - independent of using some different rotating and mirrowing base - is like dancing some 90's techno: As long as you do the same amount of steps independently of the performed sequence, even in 2-D and 3-D dimensions, the expected magic wavelets can bring in, will be there, and stays the same! For example, the result in values, and e.g the compression rates will stay exactely the same. Only all intermediate performed results (or intermediate hilbert spaces) will be different, if someone else dances differently to otherones.
For a quick test, pull the repository and then: ant && ant test. This builds a JWave.jar and the corresponding unit tests. Afterwards all units test are executed.
For a lot of own stuff with JWave, have a look at the main junit test file / method: a lot of examples!
For example, how to perform a (losless) data compression with over 98 % compression rate using all available wavelets is shown by the following junit test!
example for 1-D DFT:
Transform t = new Transform( new DiscreteFourierTransform( ) );
// arrTime = { r1, c1, r2, c2, ... }`
double[ ] arrTime = { 1., 1., 1., 1., 1., 1., 1., 1. };
double[ ] arrFreq = t.forward( arrTime ); // 1-D DFT forward
double[ ] arrReco = t.reverse( arrFreq ); // 1-D DFT reverse
example for 1-D, 2-D FWT:
Transform t = new Transform( new FastWaveletTransform( new Haar1( ) ) );
double[ ] arrTime = { 1., 1., 1., 1., 1., 1., 1., 1. };
double[ ] arrHilb = t.forward( arrTime ); // 1-D FWT Haar forward
double[ ] arrReco = t.reverse( arrHilb ); // 1-D FWT Haar reverse
double[ ][ ] matTime = { { 1., 1., 1., 1. },
{ 1., 1., 1., 1. },
{ 1., 1., 1., 1. },
{ 1., 1., 1., 1. } };
double[ ][ ] matHilb = t.forward( matTime ); // 2-D FWT Haar forward
double[ ][ ] matReco = t.reverse( matHilb ); // 2-D FWT Haar reverse
// example in 3-D in common to 2-D using a N^3 double[ ][ ][ ] space.
example for 1-D, 2-D WPT:
Transform t = new Transform( new WaveletPacketTransform( new Haar1( ) ) );
double[ ] arrTime = { 1., 1., 1., 1., 1., 1., 1., 1. };
double[ ] arrHilb = t.forward( arrTime ); // 1-D WPT Haar forward
double[ ] arrReco = t.reverse( arrHilb ); // 1-D WPT Haar reverse
double[ ][ ] matTime = { { 1., 1., 1., 1. },
{ 1., 1., 1., 1. },
{ 1., 1., 1., 1. },
{ 1., 1., 1., 1. } };
double[ ][ ] matHilb = t.forward( matTime ); // 2-D WPT Haar forward
double[ ][ ] matReco = t.reverse( matHilb ); // 2-D WPT Haar reverse
// example in 3-D in common to 2-D using a N^3 double[ ][ ][ ] space.
example for 1-D FWT of arbitrary length:
Transform t = new Transform(
new AncientEgyptianDecomposition(
new FastWaveletTransform(
new Haar1( ) ) ) );
double[ ] arrTime = { 1., 1., 1., 1., 1., 1., 1. }; // length = 7
double[ ] arrHilb = t.forward( arrTime ); // 1-D AED FWT Haar forward
// | 2 steps | 1 step | 0 |
// arrHilb = { 2., 0., 0., 0., 1.41421, 0., 1. };
double[ ] arrReco = t.reverse( arrHilb ); // 1-D AED FWT Haar reverse
example for 1-D WPT (WPD) of arbitrary length:
Transform t = new Transform(
new AncientEgyptianDecomposition(
new WaveletPacketTransform(
new Haar1( ) ) ) );
double[ ] arrTime = { 1., 1., 1., 1., 1., 1., 1. }; // length = 7
double[ ] arrHilb = t.forward( arrTime ); // 1-D AED WPT Haar forward
double[ ] arrReco = t.reverse( arrHilb ); // 1-D AED WPT Haar reverse
Have fun! :-)
If there are doubts, try mailing me, otherwise have fun with JWave.
JWave is distributed under the MIT License (MIT); this file is part of.
Copyright (c) 2008-2021 Christian ([email protected])
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
JWave is in version 200303.
version 200303:
version 180222:
version 160218:
version 160109:
version 160107:
version 160106: