SYNOPSIS

pnmscale scale_factor [pnmfile]

pnmscale -reduce reduction_factor [pnmfile]

pnmscale [{-xsize=cols | -width=cols | -xscale=factor}] [{-ysize=rows | -height=rows | -yscale=factor}] [pnmfile]

pnmscale -xysize cols rows [pnmfile]

pnmscale -pixels n [pnmfile]

Miscellaneous options:

-verbose -nomix

Minimum unique abbreviation of option is acceptable. You may use double hypens instead of single hyphen to denote options. You may use white space in place of the equals sign to separate an option name from its value.

DESCRIPTION

Reads a PBM, PGM, or PPM image as input, scales it by the specified factor or factors and produces a PGM or PPM image as output. If the input file is in color (PPM), the output will be too, otherwise it will be grayscale (PGM). This is true even if the input is a black and white bitmap (PBM), because the process of scaling can turn a combination of black and white pixels into a gray pixel.

If you want PBM output, use pgmtopbm to convert pnmscale's output to PBM. Also consider pbmreduce.

You can both enlarge (scale factor > 1) and reduce (scale factor < 1).

When you specify an absolute size or scale factor for both dimensions, pnmscale scales each dimension independently without consideration of the aspect ratio.

If you specify one dimension as a pixel size and don't specify the other dimension, pnmscale scales the unspecified dimension to preserve the aspect ratio.

If you specify one dimension as a scale factor and don't specify the other dimension, pnmscale leaves the unspecified dimension unchanged from the input.

If you specify the scale_factor parameter instead of dimension options, that is the scale factor for both dimensions. It is equivalent to -xscale=scale_factor -yscale=scale_factor .

Specifying the -reduce reduction_factor option is equivalent to specifying the scale_factor parameter, where scale_factor is the reciprocal of reduction_factor.

-xysize specifies a bounding box. pnmscale scales the input image to the largest size that fits within the box, while preserving its aspect ratio.

-pixels specifies a maximum total number of output pixels. pnmscale scales the image down to that number of pixels. If the input image is already no more than that many pixels, pnmscale just copies it as output; pnmscale does not scale up with -pixels.

If you enlarge by a factor of 3 or more, you should probably add a pnmsmooth step; otherwise, you can see the original pixels in the resulting image.

When the scale factor is not an integer (including all cases of scaling down), there are two ways to do the scaling. Which one pnmscale does is controlled by its -nomix option.

By default, pnmscale mixes the colors of adjacent pixels to produce output pixels that contain information from multiple input pixels. This makes the image look more like it would if it had infinite resolution. Note that it means the output may contain colors that aren't in the input at all.

But if you specify -nomix, pnmscale never mixes pixels. Each output pixel is derived from one input pixel. If you're scaling up, pixels get duplicated. If you're scaling down, pixels get omitted. Note that this means the image is rather distorted. If you scale up by 1.5 horizontally, for example, the even numbered input pixels are doubled in the output and the odd numbered ones are copied singly.

When the scale factor is an integer (which means you're scaling up), the -nomix option has no effect -- output pixels are always just N copies of the input pixels. In this case, though, consider using pamstretch instead of pnmscale to get the added pixels interpolated instead of just copied and thereby get a smoother enlargement.

pnmscale with -nomix is faster than without, but pnmenlarge is faster still. pnmenlarge works only on integer enlargements.

A useful application of pnmscale is to blur an image. Scale it down (without -nomix ) to discard some information, then scale it back up using pamstretch.

Or scale it back up with pnmscale and create a "pixelized" image, which is sort of a computer-age version of blurring.

PRECISION

pnmscale uses floating point arithmetic internally. There is a speed cost associated with this. For some images, you can get the acceptable results (in fact, sometimes identical results) faster with pnmscalefixed, which uses fixed point arithmetic. pnmscalefixed may, however, distort your image a little. See pnmscalefixed's man page for a complete discussion of the difference.

RELATED TO pnmscale…

AUTHOR

Copyright (C) 1989, 1991 by Jef Poskanzer.