If you have a large number of fragments of TeX to process,
it is best not to invoke dvi2bitmap for each one
individually. Also, the program does not (yet) allow you to specify
more than one DVI file as input. In this situation, it is best to
generate an input TeX or LaTeX file which contains all of the
text you wish to process, with one fragment per page.
For example, the following can help:
\documentclass{article}
\pagestyle{empty}
\special{dvi2bitmap default imageformat png}
\newcommand{\neweq}[1]{\vfil\break\special{dvi2bitmap outputfile #1.png}}
\begin{document}
\neweq{index-html-alpha}
$\alpha$
\neweq{index-html-aprime}
$A'=(0,\alpha,0,0)$
[...]
\end{document}
\neweq which simultaneously
forces a new page and inserts a special to name the output file. A
script which generates a file like this and then looks for the
resulting bitmaps, with known names of course, can run very
efficiently. Combining that with the techniques in Section 3.2 can work wonders.