In his answer to the question Tikz: using external images as building blocks. Stefan Kottwitz says:
The pgf way may be preferred if the document size matters and graphics are multiply used or if masking is desired.
When is there a problem regarding large document size caused be multiple inclusions of the same image? With a random photo in jpg-format I get the results below using the following test-files:
\documentclass{article}
\usepackage{graphicx}
\pagestyle{empty}
\begin{document}
\newcount\n
\n=0
\loop\ifnum\n<100 \advance\n by1
\includegraphics[width=5cm]{IMG_1823.JPG}
\newpage
\repeat
\end{document}
and
\documentclass{article}
\usepackage{pgf}
\pgfdeclareimage[width=5cm]{test}{IMG_1823.JPG}
\pagestyle{empty}
\begin{document}
\newcount\n
\n=0
\loop\ifnum\n<100 \advance\n by1
\pgfuseimage{test}
\newpage
\repeat
\end{document}
Compiling with pdfTeX-1.40.10 the resulting document sizes are (output from pdfinfo):
pgf-version:
File size: 1673589 bytes
Graphicx-version:
File size: 1667806 bytes
The photo is about 1.6 Mb:
$ du IMG_1823.JPG
1592 IMG_1823.JPG
Adding \pdfcompresslevel=0 and inspecting the contents of the pdf-file directly both methods seem (as far as I can tell) to embed the image data once only?