Image processing on bifurcation diagram to get small eps size

I'm producing bifurcation diagrams (which are normally used in nonlinear dynamics). These diagrams identify abrupt changes in topologies due to stability changes. These abrupt changes occur as one or more parameters pass through some critical value(s). An example is here:

http://en.wikipedia.org/wiki/File:LogisticMap_BifurcationDiagram.png

On the above figure, some image processing has been done so as to make the plot more visually pleasant. A bifurcation diagram usually contains hundreds of thousands of points and the resulting eps file can become very big. Also plotting points densely close together will just create a big blur and details won't be fine as in the figure given in the above link. Reducing marker size in Matlab help a little though to make the plot more aesthetically looking.

Journal submission in the LaTeX format requires that figures are to be submitted in the eps format. In my case one of such figures can result in about 6 MB in Matlab and even much more in Gnuplot.

For the example in the above figure, 100,000 x values are calculated for each r and one can imagine that the resulting eps file would be huge. The site however explains some image processing that makes the plot more visually pleasing. Can anyone explain to me stepwise how go about? I can't understand the explanation provided in the "summary" section.

Will the resulting image processing also reduce the figure size?

开发者_如何学编程

Furthermore, any tips on reducing the file size of such a huge eps figure?

Thanks a lot...

---

You can essentially use antialiasing. To do this, perform the computation and generate the histograms for a few (say 8) values of r within some desired range and with a high density of x values (you should have 8n bins for each x, for some large enough n). You then average each 8x8 (if you are using 8 values of r for your column) block of that column to get a single column of pixels which will appear smoother than the traditional plot. Go to the next group of r values and repeat.