Khoros Wavelet and Compression Toolbox

http://www.c3.lanl.gov/~cjhamil/Wavelets/main.html (Einblicke ins Internet, 10/1995)

The Computer Research and Applications Group (CIC/C-3) at Los Alamos National Laboratory is currently working on several projects in the area of wavelet transforms and data compression.

One of the current efforts is to develop a Khoros toolbox that implements the discrete wavelet transformation (DWT), Compression algorithms and associated support routines. Researchers at Los Alamos National Laboratory have been using Khoros as a platform in which to install programs for a variety of applications. The wavelet toolbox provides programs for both 1D and 2D wavelet transforms. The user can select the type of wavelet filter desired and interactively decide on the decomposition to be used.

Support programs allow wavelet bands to be zeroized, extracted and inserted back into the data. Statistics can be generated for each wavelet band.

A compression program has been developed to take advantage of the multiresolution decomposition derived from the wavelet transform. This program codes each wavelet subband using vector quantization.

Wavelet Toolbox

This image shows the wavelet toolbox as it appears in cantata. Clicking on the image will bring up an expanded view of the toolbox window.

The following programs are contained in the wavelet toolbox.

wvfwd - Forward wavelet transform.
wvinv - Inverse wavelet transform.
wvfilter - Select a wavelet filter to be used with the wavelet transform.
wvgentree - Generate an Octave or Uniform tree structure for the wavelet decomposition.
wvtree - This is an interactive tree structure creation program.
wxtree - This is similar to wvtree except is uses an interactive Motif program for creating the tree structure. The following image shows a view of this program.
wvinfo - Gives information about the tree structure and filters used for a wavelet transformed signal or image.
wvstats - Generates statistics such as mean value, RMS value, standard deviation, variance, skewness, kurtosis, minimum value, maximum value, and integral under the data for each wavelet subband.
wvgetband - Extract a single wavelet subband.
wvsetband - Insert data back into the wavelet transformed image or signal.
wvzero - Zero any wavelet subbands.
wvstack - Resize all wavlet subbands to be the same size and stack them into separate Khoros VIFF data bands.
wvvisual - Scales each wavelet subband so that they can be displayed. The following image shows the visualization of a wavelet transformed image using a 2 level octave tree.
wvsq - Compress a wavelet transformed image.
wvsqi - Decompress the compressed wavelet transformed image.

Compression Using Wavelet Decomposition

The following images are examples of using different compression rates on a source image. The images are as follows: Original image, compression rate of 11.7 to 1, compression rate of 14.8 to 1 and compression rate of 20 to 1.

The size of each of these images in bytes is: 262144, 22396, 17687 and 13093.

Last updated May 5, 1994.

Calvin J. Hamilton (cjhamil@lanl.gov)