Conversion from PBM bitmap format to PostScript

Introduction

This describes my efforts to submit some old PPPL reports to http://arxiv.org. The text for these were written in TeX or LaTeX, but at the time the reports were written there was no provision to include graphics in TeX output. In those days, figures were submitted to journals as glossy photographs. For the arxiv submissions, I therefore scanned the figures from paper copies of the reports and reran LaTeX including the scanned figures. The discussion here supplements the advice given at http://arxiv.org/help/bitmap.

Black & white or gray?

I initially assumed that scanning the figures as gray-scale images would be preferable. After all, that was the way to avoid the “jaggies”. However, I eventually persuaded myself that black & white images were better for two reasons:

1. Black & white scans at 300 dots per inch result in a legible printed version.
2. These can be compressed really well as PostScript (as we shall see).
3. The various tools that produce PDF files have a nasty habit of using lossy compression for gray and color images (which give a horrible blotched effect with line drawings), but always use lossless compression for bitmaps.
4. Black & white scans result in less “background” noise, e.g., from the printing on the other side of the paper.
5. Printers do not print grays consistently. Thus you might be happy with a gray image as printed on one printer, but find that the apparently “white” background prints with gray splotches on another printer.

An aside: pnmrotate uses 3 shears to implement the rotation and the manual page credits the method to

Alan W. Paeth,
“A Fast Algorithm for General Raster Rotation”,
Proc. Graphics Interface '86, Canadian Information Processing Soc., pp. 77–81, May 1986.

Useful formats

• PBM, raw bitmap format of netpbm package.
• PS, PostScript—can be inserted as graphics by LaTeX.
• PNG, lossless compressed image format, one of the standard image formats for the web—can be inserted as graphics by pdflatex.
• PDF, portable document format—can be inserted as graphics by pdflatex.

Here is a test on the bitmap image test.png (this is figure 8 from arxiv:nlin.CD/0501023). This image is a 2021 × 1956 scan at 300 dpi of a paper copy of the figure. The following table shows the size of the image file in various formats.

The PNG file is produced with

res=300
resm=`expr '(' 10000 '*' $res + 127 ')' / 254` # dots/meter
pnmtopng -compress 9 -size "$resm $resm 1" test.pbm > test.png

res=300
pnmtops -dpi $res -equalpixels -noturn -rle -nosetpage test.pbm |
  epstopdf --filter > test.pdf

In order to include graphics with latex, it is necessary to convert the image to PostScript and here we have several options.

• pnmtops is part of the netpbm package. -rle turns on run length encoding.
• imgtops is an program written by Doug Zongker and available at http://imgtops.sourceforge.net/. Version 1.0 has a “bug” where it treats a bitmap as an 8-bit image. This is fixed in the CVS repository (changes dated 2005-01-01), which I call version 1.0a. -3 emits Level 3 PostScript which will not print correctly on Level 2 PostScript printers and this is therefore not a suitable format for distributing graphics.
• tiff2ps is part of the LibTIFF package.

Recommendations

• Scan images as bitmaps at 300 dpi.
• For more flexibility, scan in gray scale and then convert to a bitmap with pgmtopbm -threshold.
• Convert bitmap to PDF format for inclusion of graphics with pdflatex.
• Convert bitmap to PS format (using pbmtops script given above) for inclusion of graphics with latex and for submission of graphics to arxiv.org.