Pattern suggestion for processing similar image types

Hello.

Some of you may have knowledge of the PGM [1] family of image formats.
As one of the activities of the Pattern Recognition course that I am
undertaking this semester, I had to write a simple tool which is capable
of parsing PGM images and realize some kinds of operations with two
given files.

The tool is supposed to be able to handle at least two variants of the
format: the /raw/ one and the /plain/ one. Both variants have the same
header structure; the difference between them lies in the "magic number"
used at the beginning of the file ('P2' for plain and 'P5' for raw) and
the way the pixel information is stored. For the /raw/ format, pixels
are stored as a sequence of 8-bit byte values; the /plain/ format, on
the other hand, keeps pixel values as whitespace-separated ASCII decimal
numbers.

What I have so far is a PGM class which is able to parse and extract
information from an input stream. The constructor of the class currently
reads the magic number and the other header data from the input stream.
Then, based on that magic number, it decides how the pixels should be
treated (as binary or ASCII values).

Although the tool is doing what it is supposed to do, I would like to
experiment with some design improvements. What pattern or patterns could
I employ in order to handle the above situation? Again, the different
formats are recognized by the "magic number"; the rest of the header is
processed identically for the formats; finally, the pixel data may be
treated in two different ways, based on the "magic number".

I would appreciate any ideas or suggestions.

[1] http://netpbm.sourceforge.net/doc/pgm.html

Thank you,

--
Ney André de Mello Zunino