Netpbm is a package of graphics programs and a programming library.
There are over 220 separate programs in the package, most of which have "pbm", "pgm", "ppm", or "pnm" in their names. For example, pnmscale and giftopnm.
For example, you might use pnmscale to shrink an image by 10%. Or use pnmcomp to overlay one image on top of another. Or use pbmtext to create an image of text. Or reduce the number of colors in an image with pnmquant.
All of the programs work with a set of graphics formats called the "netpbm" formats. Specifically, these formats are pbm, pgm, ppm, and pam. The first three of these are sometimes known generically as "pnm". Many of the Netpbm programs convert from a Netpbm format to another format or vice versa. This is so you can use the Netpbm programs to work on graphics of any format. It is also common to use a combination of Netpbm programs to convert from one non-Netpbm format to another non-Netpbm format. Netpbm has converters for about 100 graphics formats, and as a package Netpbm lets you do more graphics format conversions than any other computer graphics facility.
The Netpbm formats are all raster formats, i.e. they describe an image as a matrix of rows and columns of pixels. In the PBM format, the pixels are black and white. In the PGM format, pixels are shades of gray. In the PPM format, the pixels are in full color. The PAM format is more sophisticated. A replacement for all three of the other formats, it can represent matrices of general data including but not limited to black and white, grayscale, and color images.
Programs designed to work with PBM images have "pbm" in their names. Programs designed to work with PGM, PPM, and PAM images similarly have "pgm", "ppm", and "pam" in their names.
All Netpbm programs designed to read PGM images see PBM images as if they were PGM too. All Netpbm programs designed to read PPM images see PGM and PBM images as if they were PPM. See Implied Format Conversion.
Programs that have "pnm" in their names read PBM, PGM, and PPM but unlike "ppm" programs, they distinguish between them and their function depends on the format. For example, pnmtopng creates a black and white PNG output image if its input is PBM or PGM, but a color PNG output image if its input is PPM. And pnmscale produces an output image of the same format as the input. A hypothetical ppmscale program would also read all three PNM input formats, but would see them all as PPM and would always generate PPM output.
Programs that have "pam" in their names read all the Netpbm formats: PBM, PGM, PPM, and PAM. They sometimes treat them all as if they are PAM, using an implied conversion, but often they recognize the individual formats and behave accordingly, like a "pnm" program does. See Implied Format Conversion.
If it seems wasteful to you to have three separate PNM formats, be aware that there is a historical reason for it. In the beginning, there were only PBMs. PGMs came later, and then PPMs. Much later came PAM, which realizes the possibility of having just one aggregate format.
The formats are described in the specifications of pbm, pgm, ppm, and pam.
A program that uses the PGM library subroutines to read an image can read a PBM image as well as a PGM image. The program sees the PBM image as if it were the equivalent PGM image, with a maxval of 255. note: This sometimes confuses people who are looking at the formats at a lower layer than they ought to be because a zero value in a PBM raster means white, while a zero value in a PGM raster means black.
A program that uses the PPM library subroutines to read an image can read a PGM image as well as a PPM image and a PBM image as well as a PGM image. The program sees the PBM or PGM image as if it were the equivalent PPM image, with a maxval of 255 in the PBM case and the same maxval as the PGM in the PGM case.
A program that uses the PAM library subroutines to read an image can read a PBM, PGM, or PPM image as well as a PAM image. The program sees a PBM image as if it were the equivalent PAM image with tuple type BLACKANDWHITE. It sees a PGM image as if it were the equivalent PAM image with tuple type GRAYSCALE. It sees a PPM image as if it were the equivalent PAM image with tuple type RGB. But the program actually can see deeper if it wants to. It can tell exactly which format the input was and may respond accordingly. For example, a PAM program typically produces output in the same format as its input.
In many graphics formats, there's a means of indicating that certain parts of the image are wholly or partially transparent, meaning that if it were displayed "over" another image, the other image would show through there. Netpbm formats deliberately omit that capability, since their purpose is to be extremely simple.
In Netpbm, you handle transparency via a transparency mask in a separate (slightly redefined) PGM image. In this pseudo-PGM, what would normally be a pixel's intensity is instead an opaqueness value. See pgm. pnmcomp is an example of a program that uses a PGM transparency mask.
Another means of representing transparency information has recently developed in Netpbm, using PAM images. In spite of the argument given above that Netpbm formats should be too simple to have transparency information built in, it turns out to be extremely inconvenient to have to carry the transparency information around separately. This is primarily because Unix shells don't provide easy ways to have networks of pipelines. You get one input and one output from each program in a pipeline. So you'd like to have both the color information and the transparency information for an image in the same pipe at the same time.
For that reason, some new (and recently renovated) Netpbm programs recognize and generate a PAM image with tuple type RGB_ALPHA or GRAYSCALE_ALPHA, which contains a plane for the transparency information. See the PAM specification.
The Netpbm programs are generally useful run by a person from a command shell, but are also designed to be used by programs. A common characteristic of Netpbm programs is that they are simple, fundamental building blocks. They are most powerful when stacked in pipelines. Netpbm programs do not use graphical user interfaces (in fact, none of them display graphics at all, except for a very simple Linux Svgalib displayer) and do not seek input from a user.
Each of these programs has its own manual, as linked in the directory below.
The Netpbm programs can read and write files greater than 2 GiB wherever the underlying system can. There may be exceptions where the programs use external libraries (The JPEG library, etc.) to access files and the external library does not have large file capability. Before Netpbm 10.15 (April 2003), no netpbm program could read a file that large.
There are a few options that are present on all programs that are based on the Netpbm library, including virtually all Netpbm programs. These are not mentioned in the individual manuals for the programs.
This option was introduced in Netpbm 10.10 (October 2002).
Here is a complete list of all the Netpbm programs:
The Netpbm programming library, libnetpbm, makes it easy to write programs that manipulate graphic images. Its main function is to read and write files in the Netpbm formats, and because the Netpbm package contains converters for all the popular graphics formats, if your program reads and writes the Netpbm formats, you can use it with any formats.
But the library also contain some utility functions, such as character drawing and RGB/YCrCb conversion.
The library has the conventional C linkage. Virtually all programs in the Netpbm package are based on the Netpbm library.
As a collection of primitive tools, the power of Netpbm is multiplied by the power of all the other unix tools you can use with them. These notes remind you of some of the more useful ways to do this. Often, when people want to add high level functions to the Netpbm tools, they have overlooked some existing tool that, in combination with Netpbm, already does it.
Often, you need to apply some conversion or edit to a whole bunch of files.
As a rule, Netpbm programs take one input file and produce one output file, usually on Standard Output. This is for flexibility, since you so often have to pipeline many tools together.
Here is an example of a shell command to convert all your of PNG files (named *.png) to JPEG files named *.jpg:
Or you might just generate a stream of individual shell commands, one per file, with awk or perl. Here's how to brighten 30 YUV images that make up one second of a movie, keeping the images in the same files:
The tools find (with the -exec option) and xargs are also useful for simple manipulation of groups of files.
Some shells' "process substitution" facility can help where a non-Netpbm program expects you to identify a disk file for input and you want it to use the result of a Netpbm manipulation. Say the hypothetical program printcmyk takes the filename of a Tiff CMYK file as input and what you have is a PNG file abc.png. Try:
It works in the other direction too, if you have a program that makes you name its output file and you want the output to go through a Netpbm tool.
Netpbm contains primitive building blocks. It certainly is not a complete graphics library.
The first thing you will want to make use of any of these tools is a viewer. (On GNU/Linux, you can use ppmsvgalib in a pinch, but it is pretty limiting). zgv is a good full service viewer to use on a GNU/Linux system with the SVGALIB graphics display driver library. You can find zgv at ftp://ftp.ibiblio.org/pub/Linux/apps/graphics/viewers/svga .
zgv even has a feature in it wherein you can visually crop an image and write an output file of the cropped image using pamcut. See the -s option to zgv.
For the X inclined, there is also xzgv. See ftp://metalab.unc.edu/pub/Linux/apps/graphics/viewers/X.
xloadimage and its extension xli are also common ways to display a graphic image in X.
ImageMagick is like a visual version of Netpbm. Using the X/Window system on Unix, you can do basic editing of images and lots of format conversions. The package does include at least some non-visual tools. convert, mogrify, montage, and animate are popular programs from the ImageMagick package. ImageMagick runs on Unix, Windows, Windows NT, Macintosh, and VMS.
The Gimp is a visual image editor for Unix and X, in the same category as the more famous, less capable, and much more expensive Adobe Photoshop, etc. for Windows. See http://www.gimp.org.
To play mpeg movies, such as produced by ppmtompeg, try xine.
The free Acrobat Reader from Adobe displays PDF and converts to Postscript. You can sort of read PDF files with Netpbm by using Pstopnm. Pstopnm uses Ghostscript to read what is usually a Postscript file, but because Ghostscript also understands PDF, Pstopnm also converts from PDF to PNM. But Acrobat Reader may do it better. The Acrobat Reader for unix has a program name of "acroread" and the -toPostScript option (also see the -level2 option) is useful.
Other software from Adobe, available for purchase, interprets and creates Postscript and PDF files better than Netpbm can. "Distill" is a program that converts Postscript to PDF.
Here's some software to work with IOCA (Image Object Content Architecture): ImageToolbox ($2500, demo available). This can convert from TIFF -> IOCA and back again. Ameri-Imager ($40 Windows only).
Two packages that produce more kinds of Encapsulated Postscript than the Netpbm programs, including compressed kinds, are bmeps and imgtops.
The file program looks at a file and tells you what kind of file it is. It recognizes most of the graphics formats with which Netpbm deals, so it is pretty handy for graphics work. Netpbm's anytopnm program depends on file. See ftp://ftp.astron.com/pub/file.
The Utah Raster Toolkit serves a lot of the same purpose as Netpbm, but without the emphasis on format conversions. This package is based on the RLE format, which you can convert to and from the Netpbm formats. http://www.cs.utah.edu/research/projects/alpha1/urt.html gives some information on the Utah Raster Toolkit, but does not tell where to get it.
There are some Netpbm-like graphics tools distributed by the Army High Performance Computing Research Center at http://www.arc.umn.edu/gvl-software/media-tools.html. These operate directly on non-Netpbm format images, so they aren't included in the Netpbm package. However, you can use them with any image format by using the Netpbm format converters.
Ivtools is a suite of free X Windows drawing editors for Postscript, Tex, and web graphics production, as well as an embeddable and extendable vector graphic shell. It uses the Netpbm facilities. See http://www.ivtools.org.
Ilib is a C subroutine library with functions for adding text to an image (as you might do at a higher level with pbmtext, pnmcomp, etc.). It works with Netpbm input and output. Find it at http://www.radix.net/~cknudsen/Ilib. Netpbm also includes character drawing functions in the libnetpbm library, but they do not have as fancy font capabilities (see ppmlabel for an example of use of the Netpbm character drawing functions).
GD is a library of graphics routines that is part of PHP. It has a subset of Netpbm's functions and has been found to resize images more slowly and with less quality.
pnm2ppa converts to HP's "Winprinter" format (for HP 710, 720, 820, 1000, etc). It is a superset of Netpbm's pbmtoppa and handles, notably, color. However, it is more of a printer driver than a Netpbm-style primitive graphics building block. See http://sourceforge.net/project/?group_id=1322.
The program morph morphs one image into another. It uses Targa format images, but you can use tgatoppm and ppmtotga to deal with that format. You have to use the graphical (X/Tk) Xmorph to create the mesh files that you must feed to morph. morph is part of the Xmorph package. See http://www.colorado-research.com/~gourlay/software/Graphics/Xmorph.
To create an animated GIF, or extract a frame from one, use gifsicle. gifsicle converts between animated GIF and still GIF, and you can use ppmtogif and giftopnm to connect up to all the Netpbm utilities. See http://www.lcdf.org/gifsicle.
To convert an image of text to text (optical character recongition - OCR), use gocr (think of it as an inverse of pbmtext). See http://altmark.nat.uni-magdeburg.de/~jschulen/ocr/.
http://schaik.com/pngsuite contains a PNG test suite -- a whole bunch of PNG images exploiting the various features of the PNG format.
Another version of pnmtopng/pngtopnm is at http://www.schaik.com/png/pnmtopng.html. The version in Netpbm was actually based on that package a long time ago, and you can expect to find better exploitation of the PNG format, especially recent enhancements, in that package. It may be a little less consistent with the Netpbm project and less exploitive of recent Netpbm format enhancements, though.
jpegtran Does some of the same transformations as Netpbm is famous for, but does them specifically on JPEG files and does them without loss of information. By contrast, if you were to use Netpbm, you would first decompress the JPEG image to Netpbm format, then transform the image, then compress it back to JPEG format. In that recompression, you lose a little image information because JPEG is a lossy compression. Of course, only a few kinds of lossless transformation are possible. jpegtran comes with the Independent Jpeg Group's (http://www.ijg.org) JPEG library.
Some tools to deal with EXIF files (see also Netpbm's jpegtopnm and pnmtojpeg): To dump (interpret) an EXIF header: Exifdump ((http://topo.math.u-psud.fr/~bousch/exifdump.py) or Jhead (http://www.sentex.net/~mwandel/jhead).
A Python EXIF library and dumper: http://pyexif.sourceforge.net.
Latex2html converts Latex document source to HTML document source. Part of that involves graphics, and Latex2html uses Netpbm tools for some of that. But Latex2html through its history has had some rather esoteric codependencies with Netpbm. Older Latex2html doesn't work with current Netpbm. Latex2html-99.2beta8 works, though.
People never seem to tire of inventing new graphics formats, often completely redundant with pre-existing ones. Netpbm cannot keep up with them. Here is a list of a few that we know Netpbm does not handle (yet).
Netpbm has a long history, starting with Jef Poskanzer's Pbmplus package in 1988. The file HISTORY in the Netpbm source code contains a historical overview as well as a detailed history release by release.
Netpbm is based on the Pbmplus package by Jef Poskanzer, first distributed in 1988 and maintained by him until 1991. But the package contains work by countless other authors, added since Jef's original work. In fact, the name is derived from the fact that the work was contributed by people all over the world via the Internet, when such collaboration was still novel enough to merit naming the package after it.
Bryan Henderson has been maintaining Netpbm since 1999. In addition to packaging work by others, Bryan has also written a significant amount of new material for the package.