post_analysis.0.txt

Deviations From Previous Analysis
=================================
There is no "-l" option anymore. It's always as if "-l" was specified now.
Specifying a width and height now indicates to the program that you'd like it
to use at most that much width and/or height, but it if I needs to, it'll go
over it. For example, maybe you've specified a suggested minimum height of 300,
but you have a 340 pixels image. In this case, the program will go overboard on
the height.

There is no BMP output anymore. BMP images don't support transparency, and it's
highly desirable that the output image keeps transparency information from the
input images. Now, a PNG output is used. The "-o" option now specifies a PNG
output.

As soon as I've found I needed something with transparency, I changed
pre_analysis.txt to say png instead of bmp, but I've corrected that already.
My original thoughts involved BMP output through the SDL_SaveBMP function.

I predirected a RegionsCVSOutput module. This wasn't really used. That module
was supposed to output CSV through a SDL RWOps structure. What was actually
done is a function in the Main module which outputs the CSV result to a file.
Much simpler, and just as good.

BMP Output
==========
BMP output seems useful still. It should be easy to do it. Binary transparency
can be faked through the use of a background color that doesn't appear in any
input image. That background color would be another output of the program.

PNG Output
==========
I don't understand about PNG images processing. The code in xPNG.c and xPNG.h
was mostly a copy/paste from the web. I tweaked with the code a little so it
could do better error reporting, but that's about it.