$ pamdice myimage.ppm -outstem=myimage_part -width=10 -height=8
$ pamundice myimage_part_%1d_%1a.ppm -across=10 -down=8 >myimage.ppm
$ pamundice myimage_part_%2a -across=13 -hoverlap=9
See pamdice for more examples.
You can use the minimum unique abbreviation of the options. You can use two hyphens instead of one. You can separate an option name from its value with white space instead of an equals sign.
This program is part of Netpbm.
pamundice reads a bunch of Netpbm images as input and combines them as a grid of tiles into a single output image of the same kind on Standard Output.
You can optionally make the pieces overlap.
The images can either be in files whose names indicate where they go in the output (e.g. "myimage_part_03_04" could be the image for Row 3, Column 4 - see the input_filename_pattern argument) or listed in a file, with a -listfile option.
The input images must all have the same format (PAM, PPM, etc.) and maxval and for PAM must have the same depth and tuple type. All the images in a rank (horizontal row of tiles) must have the same height. All the images in a file (vertical column of tiles) must have the same width. But it is not required that every rank have the same height or every file have the same width.
pamdice is the inverse of pamundice. You can use pamundice to reassemble an image sliced up by pamdice. You can use pamdice to recreate the tiles of an image created by pamundice, but to do this, the original ranks must all have been the same height except for the bottom one and the original files must all have been the same width except the right one.
One use for this is to process an image in pieces when the whole image is too large to process. For example, you might have an image so large that an image editor can't read it all into memory or processes it very slowly. You can split it into smaller pieces with pamdice, edit one at a time, and then reassemble them with pamundice.
Of course, you can also use pamundice to compose various kinds of checkerboard images, for example, you could write a program to render a chessboard by computing an image of each square, then using pamundice to assemble them into a board.
An alternative to join images in a single direction (i.e. a single rank or a single file) is pamcat. pamcat provides alternative ways to identify the input images: you can supply them serially on Standard Input or list them by name as command line arguments.
To join piecewise photographs, use pnmstitch instead of pamundice, because it figures out where the pieces overlap, even if they don't overlap exactly vertically or horizontally.
To create an image of the same tile repeated in a grid, that's pnmtile.
pnmindex does a similar thing to pamundice: it combines a bunch of small images in a grid into a big one. But its purpose is to produce an index image of the input images. So it leaves space between them and has labels for them, for example.
Unless you use a -listfile option, there is one non-option argument, and it is mandatory: input_filename_pattern. This tells pamundice what files contain the input tiles.
pamundice reads the input images from files which are named with a pattern that indicates their positions in the combined image. For example, tile_00_05.ppm could be the 6th tile over in the 1st rank, while tile_04_01 is the 2nd tile over in the 5th rank.
(But see -indexfile, which can cause the tiles to be rearranged so that the file whose name indicates it is the 6th tile over in the 1st rank could actually be placed by pamundice in any arbitrary position in its output).
You cannot supply any of the data on Standard Input, and the files must be the kind that pamundice can close and reopen and read the same image a second time (e.g. a regular file is fine; a named pipe is probably not).
input_filename_pattern is a printf-style pattern. (See the standard C library printf subroutine). For the example above, it would be tile_%2d_%2a.ppm. The only possible conversion specifiers are:
The number between the % and the conversion specifier is the precision and is required. It says how many characters of the file name are described by that conversion. The rank or file number is filled with leading zeroes as necessary.
So the example tile_%2d_%2a.ppm means to get the name of the file that contains the tile at Rank 0, File 5, you:
Note that this pattern describes file names that pamdice produces, except that the precision may be more or less. (See the -numberwidth option of pamdice).
In addition to the options common to all programs based on libnetpbm (most notably -quiet, see Common Options), pamundice recognizes the following command line options:
You cannot specify this together with -indexfile.
Default is 1.
You cannot specify this together with -indexfile.
Default is 1.
There must not be any input image narrower than this.
Note that this undoes the effect of the same -hoverlap option of pamdice.
Default is zero -- no overlap.
The named file contains file names, one per line. Each file contains the image for one tile, in row-major order, top to bottom, left to right. So the first file is the upper left tile, the second is the one to right of that, etc. The number of lines in the file must be equal to the number of tiles in the output, the product of the -across and -down values.
The file names have no meaning to pamundice. You can list the same file multiple times to have identical tiles in the output.
If you create the tile files with pamdice, you can use the -listfile option of pamdice to produce a suitable list file for reassembling the tiles into the same positions they had in the original large image.
You cannot specify this together with -indexfile.
This option was new in Netpbm 10.90 (March 2020).
A typical way to generate an index image to use with pamundice is to use the -indexfile option of pamdice, and then possibly manipulate the resulting index image with other Netpbm tools.
While the sample values in the index image are defined to be rank and file positions in some original large image, and so are documented in those terms, in practice, pamundice doesn't assign any meaning to the numbers -- they are just substitutions for two variables in the input file name pattern (given as a program argument).
The maxval of the index image is irrelevant, except that no sample in the index image can be greater than it.
The depth of the index file must be at least two (rank and file), but the program ignores any additional planes.
The program ignores the tuple type.
If you don't specify this option, the position of a tile in the ouput is given directly by the rank and file portions of the tile file name, or if you specify -listfile, by order in which the tile files are listed in the list file.
You cannot specify this together with -listfile, -down, or -across.
This option was new in Netpbm 11.10 (March 2025).
Here is an example of using a list file:
Example usage of -listfile:
pamdice -width=100 -height=100 -outstem=img_part -listfile=listfile \
img.ppm
pamundice -listfile=listfile -across=6 -down=4 > img_copy.ppm
Note that in the above example, you have to know somehow that you got 6 files and 4 ranks from pamdice and then specify the correct -across and -down options on pamundice. A slight error ruins the process. A better way is to use -indexfile:
pamdice -width=100 -height=100 -outstem=img_part \
-numberwidth=3 -indexfile=index.pam img.pam
pamundice -indexfile=index.pam img_part_%3d_%3a.pam > img_copy.pam
Here is an example of flipping a large image. pamflip often requires the entire image to be in real memory at once, which can be impossible or impractical with a very large image. This method does it in smaller pieces so as to require less memory:
Example of -indexfile with pamflip operation:
pamdice -width=2500 -height=2500 -outstem=tile \
-numberwidth=3 -indexfile=index.pam hugeImage.ppm
# flip every tile
for tile in tile_???_???.ppm
do
pamflip -r180 $tile > ${tile%.ppm}.flipped.ppm
done
# flip indexfile - Simply apply the same flip operation
pamflip -r180 index.pam > index_flipped.pam
pamundice -indexfile=index_flipped.pam tile_%3d_%3a.flipped.ppm \
> hugeImage_flipped.ppm
pamundice was new in Netpbm 10.39 (June 2007). Before that, pamcat is the best substitute.