Man1 - pamtotiff.1

By default, the output goes to Standard Output. Alternatively, you can specify an output file with the -output option and pamtotiff will write its output directly to that file.

Because of the way the TIFF library (which pamtotiff uses) works, when the program writes to Standard Output, it generates the entire TIFF image in a temporary file and then copies it to Standard Output; you may see negative performance effects of this.

The -output method avoids the performance effects of the copy through the temporary file, but there are restrictions on the output file: it must be seekable and it must be readable. The program fails if it is not. With Standard Output, neither of those restrictions applies.

With -output, if the file already exists and has data in it, pamtotiff appends the image to the existing TIFF file. (A TIFF file may contain multiple images).

By default, pamtotiff creates the file named by -output if it does not already exist. But if you also specify -append, the program fails if the file named by -output does not already exist.

Before Netpbm 10.67 (June 2014), there is no -output option and Standard Output must be seekable. So pipes are out.

Before Netpbm 10.67 (June 2014), you could append to Standard Output. See below. The current program does not have the ability; you must use -output to append to an existing TIFF file.

The difference above means current pamtotiff is actually not backward compatible, which is a rare thing for Netpbm. But it’s a good thing because the previous function was very strange and probably hardly ever exploited.

Old Versions

As alluded to above, pamtotiff does output very differently in releases before 10.67. The following describes the old function, which is unusual both for Netpbm and for Unix programs in general.

• The output file must be seekable. pamtotiff does not write it sequentially. Therefore, you can’t use a pipe; you can’t pipe the output of pamtotiff to some other program. But any regular file should work.
• If the output file descriptor is readable, you must either specify -append so as to add to the existing file, or make sure the file is empty. Otherwise, pamtotiff will fail with an unhelpful message telling you that a TIFF library function failed to open the TIFF output stream.
• If you are converting multiple images (your input stream contains multiple images), the output file must be both readable and writable.

If you’re using a Unix command shell to run pamtotiff, you use facilities of your shell to set up Standard Output. In Bash, for example, you would set up a write-only Standard Output to the file /tmp/myimage.tiff like this:

      $ pamtotiff myimage.pnm >/tmp/myimage.tiff

In Bash, you would set up a read/write Standard Output to the file /tmp/myimage.tiff like this:

      $ pamtotiff myimage.pnm 1<>/tmp/myimage.tiff

pamtotiff uses the Libtiff.org TIFF library (or whatever equivalent you provide) to generate the TIFF output. Details of the format it produces are therefore controlled by that library.

By default, pamtotiff creates a TIFF file with no compression. This is your best bet most of the time. If you want to try another compression scheme or tweak some of the other even more obscure output options, there are a number of options which to play.

Before Netpbm 8.4 (April 2000), the default was to use LZW compression. But then new releases of the TIFF library started omitting the LZW compression capability because of concern about patents on LZW. So since then, the default has been no compression. The LZW patents have now expired and new TIFF libraries do LZW, but the pamtotiff behavior remains the same for compatibility with older TIFF libraries and applications of pamtotiff.

The -none, -packbits, -lzw, -g3, -g4, -flate, and -adobeflate options are used to override the default and set the compression scheme used in creating the output file.

The -predictor option is meaningful only with LZW compression: a predictor value of 2 causes each scanline of the output image to undergo horizontal differencing before it is encoded; a value of 1 forces each scanline to be encoded without differencing. By default, pamtotiff creates a TIFF file with msb-to-lsb fill order. The -msb2lsb and -lsb2msb options are used to override the default and set the fill order used in creating the file.

With some older TIFF libraries, -lzw doesn’t work because the TIFF library doesn’t do LZW compression. This is because of concerns about Unisys’s patent on LZW which was then in force. Actually, with very old TIFF libraries, -lzw works because no distributors of the TIFF library were sensitive yet to the patent issue.

-flate chooses ’flate’ compression, which is the patent-free compression common in the Unix world implemented by the ’Z’ library. It is what the PNG format uses.

Fax Compression

If you have bilevel data (e.g. PBM), you can generate a TIFF that uses the same compression scheme specified for use by fax machines. See the Fax*Format*(1) page for more information on these compression schemes.

These formats all relate to ITU Group 3 and Group 4 fax machine standards.

The -g3 option chooses MH or MR compression: MR with the additional option -2d; MH without it. -g4 selects MMR. The option names are a little unfortunate and historical, but are consistent with the TIFF specification.

MMR has a better compression ratio than the other two.

-fill specifies that for MH or MR compression, each encoded scanline shall be zero-filled to a byte boundary.

-2d and -fill are meaningful only with -g3.

The -msb2lsb and lsb2msb options control the fill order.

The fill order is the order in which pixels are packed into a byte in the Tiff raster, in the case that there are multiple pixels per byte. msb-to-lsb means that the leftmost columns go into the most significant bits of the byte in the Tiff image. However, there is considerable confusion about the meaning of fill order. Some believe it means whether 16 bit sample values in the Tiff image are little-endian or big-endian. This is totally erroneous (The endianness of integers in a Tiff image is designated by the image’s magic number). However, ImageMagick and older Netpbm both have been known to implement that interpretation. 2001.09.06.

If the image does not have sub-byte pixels, these options have no effect other than to set the value of the FILLORDER tag in the Tiff image (which may be useful for those programs that misinterpret the tag with reference to 16 bit samples).

-color tells pamtotiff to produce a color, as opposed to grayscale, TIFF image if the input is PPM, even if it contains only shades of gray. Without this option, pamtotiff produces a grayscale TIFF image if the input is PPM and contains only shades of gray, and at most 256 shades. Otherwise, it produces a color TIFF output. For PBM and PGM input, pamtotiff always produces grayscale TIFF output and this option has no effect.

The -color option can prevent pamtotiff from making two passes through the input file, thus improving speed and memory usage. See .

-truecolor tells pamtotiff to produce the 24-bit RGB form of TIFF output if it is producing a color TIFF image. Without this option, pamtotiff produces a colormapped (paletted) TIFF image unless there are more than 256 colors (and in the latter case, issues a warning).

The -truecolor option can prevent pamtotiff from making two passes through the input file, thus improving speed and memory usage. See .

The -color and -truecolor options did not exist before Netpbm 9.21 (December 2001).

If pamtotiff produces a grayscale TIFF image, this option has no effect.

The -minisblack and -miniswhite options force the output image to have a ’minimum is black’ or ’minimum is white’ photometric, respectively. If you don’t specify either, pamtotiff uses minimum is black except when using Group 3 or Group 4 compression, in which case pamtotiff follows CCITT fax standards and uses ’minimum is white.’ This usually results in better compression and is generally preferred for bilevel coding.

Before February 2001, pamtotiff always produced ’minimum is black,’ because of a bug. In either case, pamtotiff sets the photometric interpretation tag in the TIFF output according to which photometric is actually used.

The -indexbits option is meaningful only for a colormapped (paletted) image. In this kind of image, the raster contains values which are indexes into a table of colors, with the indexes normally taking less space that the color description in the table. pamtotiff can generate indexes of 1, 2, 4, or 8 bits. By default, it will use 8, because many programs that interpret TIFF images can’t handle any other width.

But if you have a small number of colors, you can make your image considerably smaller by allowing fewer than 8 bits per index, using the -indexbits option. The value is a comma-separated list of the bit widths you allow. pamtotiff chooses the smallest width you allow that allows it to index the entire color table. If you don’t allow any such width, pamtotiff fails. Normally, the only useful value for this option is 1,2,4,8, because a program either understands the 8 bit width (default) or understands them all.

In a Baseline TIFF image, according to the 1992 TIFF 6.0 specification, 4 and 8 are the only valid widths. There are no formal standards that allow any other values.

This option was added in June 2002. Before that, only 8 bit indices were possible.

There are lots of tag types in the TIFF format that don’t correspond to any information in the PNM format or to anything in the conversion process. For example, a TIFF_ARTIST tag names the artist who created the image.

You can tell pamtotiff explicitly to include tags such as this in its output with the -tag option. You identify a list of tag types and values and pamtotiff includes a tag in the output for each item in your list.

The value of -tag is the list of tags, like this example:

      -tag=subfiletype=reducedimage,documentname=Fred,xposition=25

As you see, it is a list of tag specifications separated by commas. Each tag specification is a name and a value separated by an equal sign. The name is the name of the tag type, except in arbitrary upper/lower case. One place to see the names of TIFF tag types is in the TIFF library’s tiff.h file, where there is a macro defined for each consisting of ’TIFF_’ plus the name. E.g. for the SUBFILETYPE tag type, there is a macro TIFF_SUBFILETYPE.

The format of the value specification for a tag (stuff after the equal sign) depends upon what kind of value the tag type has:

• Integer: a decimal number
• Floating point number: a decimal number
• String: a string
• Enumerated (For example, a ’subfiletype’ tag takes an enumerated value. Its possible values are REDUCEDIMAGE, PAGE, and MASK.): The name of the value. You can see the possible value names in the TIFF library’s tiff.h file, where there is a macro defined for each consisting of a qualifier plus the value name. E.g. for the REDUCEDIMAGE value of a SUBFILETYPE tag, you see the macro FILETYPE_REDUCEDIMAGE.

The TIFF format assigns a unique number to each enumerated value and you can specify that number, in decimal, as an alternative. This is useful if you are using an extension of TIFF that pamtotiff doesn’t know about.

If you specify a tag type with -tag that is not independent of the content of your PNM source image and pamtotiff’s conversion process (i.e. a tag type in which pamtotiff is interested), pamtotiff fails. For example, you cannot specify an IMAGEWIDTH tag with -tag, because pamtotiff generates an IMAGEWIDTH tag that gives the actual width of the image.

-tag was new in Netpbm 10.31 (December 2005).

See The Output File .

-output names the output file. Without this option pamtotiff writes to Standard Output.

The -append option tells pamtotiff only to append to the file named by output; not create it. Without this option, the program creates the file it does not already exist. But even then, if the file does already exist, the program appends the image to what is in the file already. (A TIFF file may contain multiple images).

-append has no effect if you don’t also specify -output.

Before Netpbm 10.67 (June 2014), -append means something rather different: it means to append the image to the output TIFF file (which is always Standard Output in 10.67) instead of replacing its contents.

-append was new in Netpbm 10.27 (March 2005).

You can use the -rowsperstrip option to set the number of rows (scanlines) in each strip of data in the output file. By default, the output file has the number of rows per strip set to a value that will ensure each strip is no more than 8 kilobytes long.

Like most Netpbm programs, pamtotiff’s function is mostly undefined if the input is PAM image with tuple type other than BLACKANDWHITE, GRAYSCALE, or RGB. Most of the statements in this manual assume the input is not such an exotic PAM. But there is a little defined processing of other PAM subformats.

pamtotiff assumes any 1 plane PAM image is BLACKANDWHITE or GRAYSCALE (and doesn’t distinguish between those two).

pamtotiff assumes a PAM with more than 1 plane is of tuple type RGB except with that number of planes instead of 3. pamtotiff doesn’t really understand red, green, and blue, so it has no trouble with a 2-component or 5-component color space. The TIFF format allows an arbitrary number of color components, so pamtotiff simply maps the PAM planes directly to TIFF color components. I don’t know if the meanings of 5 components in a TIFF image are standard at all, but the function is there if you want to use it.

Note that pamtotiff may generate either a truecolor or colormapped image with an arbitrary number of color components. In the truecolor case, the raster has that number of planes. In the colormapped case, the raster has of course 1 plane, but the color map has all the color components in it.

The most common reason for a PAM to have extra planes is when the tuple type is xxx_ALPHA, which means the highest numbered plane is a transparency plane (alpha channel). At least one user found that a TIFF with an extra plane for transparency was useful.

Note that the grayscale detection works on N-component colors, so if your planes aren’t really color components, you’ll want to disable this via the -color option.

pamtotiff reads the input image once if it can, and otherwise twice. It needs that second pass (which happens before the main pass, of course) to analyze the colors in the image and generate a color map (palette) and determine if the image is grayscale. So the second pass happens only when the input is PPM. And you can avoid it then by specifying both the -truecolor and -color options.

If the input image is small enough to fit in your system’s file cache, the second pass is very fast. If not, it requires reading from disk twice, which can be slow.

When the input is from a file that cannot be rewound and reread, pamtotiff reads the entire input image into a temporary file which can, and works from that. Even if it needs only one pass.

Before Netpbm 9.21 (December 2001), pamtotiff always read the entire image into virtual memory and then did one, two, or three passes through the memory copy. The -truecolor and -color options did not exist. The passes through memory would involve page faults if the entire image did not fit into real memory. The image in memory required considerably more memory (12 bytes per pixel) than the cached file version of the image would.

A Tiff image may contain information about the resolution of the image, which means how big in real dimensions (centimeters, etc.) each pixel in the raster is. That information is in the TIFF XRESOLUTION, YRESOLUTION, and RESOLUTIONUNIT tags. By default, pamtotiff does not include any tags of these types, but you can specify them with the -tags option.

There are also options -xresolution, -yresolution, and -resolutionunit, but those are obsolete. Before -tags existed (before Netpbm 10.31 (December 2005), they were the only way.

Note that the number of pixels in the image and how much information each contains is determined independently from the setting of the resolution tags. The number of pixels in the output is the same as in the input, and each pixel contains the same information. For your resolution tags to be meaningful, they have to consistent with whatever created the PNM input. E.g. if a scanner turned a 10 centimeter wide image into a 1000 pixel wide PNM image, then your horizontal resolution is 100 pixels per centimeter, and if your XRESOLUTION tag says anything else, something that prints your TIFF image won’t print the proper 10 centimeter image.

The value of the XRESOLUTION tag is a floating point decimal number that tells how many pixels there are per unit of distance in the horizontal direction. -yresolution is analogous for the vertical direction.

The unit of distance is given by the value of the RESOLUTIONUNIT option. That value is either INCH, CENTIMETER, or NONE. NONE means the unit is arbitrary or unspecified. This could mean that the creator and user of the image have a separate agreement as to what the unit is. But usually, it just means that the horizontal and vertical resolution values cannot be used for anything except to determine aspect ratio (because even though the unit is arbitrary or unspecified, it has to be the same for both resolution numbers).

If you don’t use a -tag option to specify the resolution tag and use the obsolete options instead, note the following:

• If you don’t include an specify -xresolution, the Tiff image does not contain horizontal resolution information. Likewise for -yresolution. If you don’t specify -resolutionunit, the default is inches.
• Before Netpbm 10.16 (June 2003), -resolutionunit did not exist and the resolution unit was always inches.