pnmcolormap - create quantization color map for a Netpbm image
This program is part of Netpbm(1).
pnmcolormap reads a PNM or PAM image as input, chooses ncolors colors to best represent the image and writes a PNM color map defining them as output. A PAM image may actually contain tuples of any kind, but pnmcolormap's concept of the tuple values that best represent the ones present in the image may not make sense if the tuple type isn't RGB or GRAYSCALE. The design of the program, and the rest of this manual, assumes the tuples represent colors.
You can use this map as input to pnmremap on the same input image to quantize the colors in that image, I.e. produce a similar image with fewer colors. pnmquant does both the pnmcolormap and pnmremap steps for you.
A PNM colormap is a PNM image of any dimensions that contains at least one pixel of each color in the set of colors it represents. The ones pnmcolormap generates have exactly one pixel of each color, except where padding is necessary with the -square option.
The quantization method is Heckbert's "median cut". See QUANTIZATION METHOD .
The output image is of the same format (PBM, PGM, PPM, PAM) as the input image. Note that a colormap of a PBM image is not very interesting.
The colormap generally has the same maxval as the input image, but pnmcolormap may reduce it if there are too many colors in the input, as part of its quantization algorithm.
pnmcolormap works on a multi-image input stream. In that case, it produces one colormap that applies to all of the colors in all of the input images. All the images must have the same format, depth, and maxval (but may have different height and width). This is useful if you need to quantize a bunch of images that will form a movie or otherwise be used together -- you generally want them all to draw from the same palette, whereas computing a colormap separately from each image would make the same color in two images map to different colors. Before Netpbm 10.31 (December 2005), pnmcolormap ignored any image after the first.
If you want to create a colormap without basing it on the colors in an input image, pamseq, ppmmake, and pnmcat can be useful.
The single parameter, which is required, is the number of colors you want in the output colormap. pnmcolormap may produce a color map with slightly fewer colors than that. You may specify all to get a colormap of every color in the input image (no quantization). When you specify all, the function is essentially the same as that of ppmhist -map. ppmhist is much older.
In addition to the options common to all programs based on libnetpbm
(most notably -quiet, see
Common Options ), pnmcolormap recognizes the following command line options:
All options can be abbreviated to their shortest unique prefix. You may use two hyphens instead of one to designate an option. You may use either white space or an equals sign between an option name and its value.
A quantization method is a way to choose which colors, being fewer in number than in the input, you want in the output. pnmcolormap uses Heckbert's "median cut" quantization method.
This method involves separating all the colors into "boxes," each holding colors that represent about the same number of pixels. You start with one box and split boxes in two until the number of boxes is the same as the number of colors you want in the output, and choose one color to represent each box.
There are three ways pnmcolormap can choose the box to split in
-splitpixelct, splitcolorct, and splitspread were new
in Netpbm 10.88 (September 2019). Before that, pnmcolormap always
splits the box containing the most pixels.
When you split a box, you do it so each sub-box has the same number of pixels (except one sub-box has more if the full box has an odd number), with the 'greatest' pixels in one sub-box and the 'least' pixels in the other. "Greater," for a particular box, means it is brighter in the color component (red, green, blue) which has the largest spread in that box. pnmcolormap gives you two ways to define "largest spread.": 1) largest spread of brightness; 2) largest spread of contribution to the luminosity of the color. E.g. red is weighted much more than blue. Select among these with the -spreadbrightness and -spreadluminosity options. The default is -spreadbrightness. Where there are multiple colors of the median magnitude, they are distributed arbitrarily among between the subboxes. This arbitrary distribution depends upon what the system's qsort function does with multiple equal values, so pnmcolormap may produce slightly different results on different systems.
pnmcolormap provides three ways of choosing a color to represent a box: 1) the center color - the color halfway between the greatest and least colors in the box, using the above definition of "greater"; 2) the mean of the colors (each component averaged separately by brightness) in the box; 3) the mean weighted by the number of pixels of a color in the image.
Select among these with the -center, -meancolor, and -meanpixel options. The default is -center.
Note that in all three methods, there may be colors in the output which do not appear in the input at all.
Also note that the color chosen to represent the colors in Box A the best
may also represent a color in Box B better than the color chosen to represent
the colors in Box B the best. This is true for various measures of goodness
of representation of one color by another. In particular, if you
use pnmremap to map the colors in the very image that you used to
create the color map to the colors in that colormap, the colors in Box B will
often map to the color pnmcolormap chose to represent some other box
and in fact the color pnmcolormap chose to represent Box B may not
appear in the pnmremap output at all.
"Color Image Quantization for Frame Buffer Display" by Paul Heckbert, SIGGRAPH '82 Proceedings, page 297.
Before Netpbm 10.15 (April 2003), pnmcolormap used a lot more memory for large images because it kept the entire input image in memory. Now, it processes it a row at a time, but because it sometimes must make multiple passes through the image, it first copies the input into a temporary seekable file if it is not already in a seekable file.
pnmcolormap first appeared in Netpbm 9.23 (January 2002). Before that, its function was available only as part of the function of pnmquant (which was derived from the much older ppmquant). Color quantization really has two main subfunctions, so Netpbm 9.23 split it out into two separate programs: pnmcolormap and pnmremap and then Netpbm 9.24 replaced pnmquant with a program that simply calls pnmcolormap and pnmremap.