NAME

SYNOPSIS

rlequant [ -b bits ] [ -c ] [ -d ] [ -f ] [ -i cubeside ] [ -m ] [ -n colors ] [ -r mapfile ] [ -o outfile ] [ infile ]

DESCRIPTION

-b bits

The colors in the input image will be "prequantized" to this many bits before applying the variance-based method. Two internal tables of size 2^(3*bits) are allocated, so values of bits greater than 6 are likely to cause thrashing or may prevent the program from running at all. The default value of bits is 5. It must be less than or equal to 8 and greater than 0.

-c

Only the color map will be output; the image will not be digitized. The output file will be a 0x0 RLE file with a color map, suitable for input to rleldmap(1), rledither(1), or rlequant -r.

-d

Floyd Steinberg dithering is performed on the output. This is very helpful for images being quantized to a small number of colors.

-f

If this option is specified, a faster approximation will be used. In most cases, the error so introduced will be barely noticeable.

-i cubeside

Initializes the output color map with a "color cube" of size cubeside^3. I.e., if -i 2 were specified, the 8 corners of the color cube (black, red, green, blue, yellow, cyan, magenta, white) would be added to the output colormap. This reduces the number of colors available for quantization. The color cube will be used to quantize the output image, but will not otherwise affect the choice of representative colors.

-m

Computes a single color map suitable for quantizing all the input images. This is useful when the quantized images will be used as a "movie" (e.g., with the -m flag of getx11(1)). The input may not come from a pipe when this option is specified, unless -c is also specified.

-n colors

The output image will be quantized to at most colors colors. It might have fewer if the input image has only a few colors itself. The default value of colors is 256 - cubesize^3 - mapsize. It must be less than or equal to 256. If a color cube (-c) or an input map (-r) is given, colors may be 0; otherwise it must be greater than 0.

-r mapfile

The color map from the RLE file mapfile will be added to the output color map. The number of colors in the input color map, mapsize is calculated as follows: If a color_map_length comment is present in mapfile, its value is used. If not, the size of the color map (usually 256) is used (the rlehdr(1) program will display the color map size and the comment, if present). The input color map will be used to quantize the output image, but will not otherwise affect the choice of representative colors. If the combination -n 0 -r mapfile is specified, then rlequant will just quantize (and dither, if requested) the input images to the given colormap. This is usually faster than using rledither.

-o outfile

The output will be written to the file outfile. If not specified, or if outfile is "-", the output will be written to the standard output stream.

infile

This file contains one or more concatenated RLE images. Each will be processed in turn. A separate quantization map will be constructed for each image. If not specified, or if infile is "-", the image(s) will be read from the standard input stream.

EXAMPLES

rlequant file.rle

Quantizes file.rle to 256 colors using a 5-bit pre-quantization. If file.rle has multiple images, each will get its own (different) colormap.

rlequant -m file.rle

Quantizes file.rle to 256 colors using a 5-bit pre-quantization. If file.rle has multiple images, they will all be used to choose the color map, and will all be quantized to the same color map.

cat *.rle | rlequant -m -c >map.rle ;

cat *.rle | rlequant -n 0 -r map.rle
Computes a single colormap based on all the images in the files *.rle, then quantizes each image to that color map. The output is the stream of quantized images.

rlequant -i 4 -d file.rle

Compute 192 representative colors for each image in file.rle, add a 4x4x4 color cube to the resulting color map, and then quantize the image to the resulting set of colors with dithering.

rlequant -b 6 file.rle

Quantize file.rle to 256 colors using a 6-bit prequantization. This provides slightly more precision in color matching than does a 5-bit prequantization. It also runs significantly slower and requires approximately 8 times the memory for its intermediate storage.

SEE ALSO

AUTHOR