 |
 |
|
|
 |
 |
 |
|||
Java Wavelet Demo
Two Scale Pick Analysis Denoise Compression Engine Wavelet based Data Compression
Wavelet based lossy compression techniques have three steps in general:
- Transform: data are first transformed into wavelet domain. This step is invertible.
- Quantization: the wavelet coefficients are quantized to a finite alphabet. This step is not invertible, thus introduces the so called quantization noise.
- Entropy Coding: the resulting symbols after quantization are further entropy coded to reduce the bit rate. This step is also invertible.
This applet shows the basic idea of wavelet based lossy compression. First you choose a signal and a wavelet. The wavelet coefficients of the signal are shown in the figure at the left hand side. You can click in the left figure to set the quantization step size. Two green horizontal lines will appear in the input figure corresponding to positive and negative values of the step size you set. The signal is transformed into the wavelet domain, and further quantized using the given step size. Then reconstructed signal will be computed and shown in the right figure.
You can change the step size by clicking and dragging mouse in the left figure. You can also change the quantization method by choosing either the uniform scalar quantizer or the deadzone quantizer.
It is also enlighting to look the wavelet coefficients, and see how they change for different quantization methods and step size. You can switch to the wavelet domain by making a choice under the right side figure. The red line is the original signal; the yellow line is the compressed and reconstructed signal.
You can also use your own signal. Simply enter the URL of the file, then choose [File/Load Signal] from the menu bar. The format of the file is very simple. Each line in the file should contain the ASCII representation of a number. The demo will only use the first 1024 data points. If the data file contains less than 1024 points, zeros will be padded to your data.
Reference
J. M. Shapiro, Embedded Image Coding Using Zerotrees of Wavelet Coefficients. IEEE Transaction on Signal Processing V41 p3445-3462, 1993.Copyright Rice University, 1996-1997.
Haitao Guo
