WAVELET(1)

NAME

wavelet - perform a Haar wavelet decomposition or reconstruction

On this page

SYNOPSIS
DESCRIPTION
OPTIONS
EXAMPLES
SEE ALSO
AUTHOR
COPYRIGHT
BUG REPORTS

SYNOPSIS

wavelet {-d | -r} [-1 | -2] [-t _datatype_] [-# _channels_] [-w _width_] [-n _scanlines_] [-s _squaresize_] [-W _out_size_] [-S _out_size_] [-R _avg_size_] [-D _level_] < datastream > wavelets

The wavelet command performs a Haar wavelet transform on a raw binary dataset read from standard input, writing the transformed result to standard output. It can either decompose a signal into successively coarser "average" terms plus "detail" coefficients, or reconstruct the original signal from such a decomposition. Decomposition and reconstruction are inverse operations.

The transform operates on interleaved, fixed-width samples, so it works directly on image formats such as pix(5) and bw(5) by treating each pixel as one sample with the appropriate number of channels. A two-dimensional transform requires a square dataset.

Because the Haar transform sums and differences sample values, results can overflow narrow integer types. For lossless decomposition and reconstruction, convert the data to the next larger data type (via the -t option) before processing.

OPTIONS

-d: Perform a decomposition.
-r: Perform a reconstruction.
-1: Perform a one-dimensional (per-scanline horizontal) transform. This is the default.
-2: Perform a two-dimensional transform (interleaved horizontal and vertical). The dataset must be square.
-t datatype: Specify the data type of the samples, which is also the type used for the wavelet arithmetic. The letter is the first character of the C type: c (char), s (short), i (int), l (long), f (float), d (double). The default is char.
-# channels: Number of values per sample point. For example, use -# 3 to transform a pix(5) file so the red, green, and blue channels are transformed individually. The default is 3.
-w width: Number of samples per scanline. The default is 512.
-n scanlines: Number of scanlines in the dataset. The default is 512.
-s squaresize: Set the width and number of scanlines simultaneously to the same value (a square dataset).
-W out_size: Width of the average image in the output. Normally a full decomposition reduces this to 1; specify a power-of-two value at which decomposition or reconstruction should stop. The output then holds an averaged image of the given size with the remaining data stored as detail coefficients.
-S out_size: Same as -W, expressed as the square size of a two-dimensional transform limit.
-R avg_size: Resume a transform from a decomposition that left off with an average image of avg_size.
-D level: Set the debug output level. Diagnostic information about data type, channels, width, height, and limit is written to standard error.

EXAMPLES

Example 1. Display a horizontal decomposition

wavelet -d -# 3 -s 512 < img.pix | pix-fb

Decomposes each scanline of a 512x512 color pix(5) image and displays the result on a framebuffer with pix-fb(1). Each scanline shows a shrinking "average" toward the left with detail coefficients to the right.

Example 2. Two-dimensional decomposition with a stopping size

wavelet -d -2 -# 3 -s 1024 -S 64 < img.pix > img.wlt

Performs a 2D decomposition of a 1024x1024 color image, stopping once the average has been reduced to 64x64. The output image holds a 64x64 version of the original in the lower-left corner with detail terms filling the remainder, ready to be inverted with wavelet -r -2 -s 1024 -R 64.