It took me a little over one year to finish my project. It involved:
I learned many things that year (even things relating to my project!) The most important lesson I learned is to NOT do a 2 credit project that attempts to prove something. My project could have been something as simple as creating a GUI interface for FTP, but nooooo, I had to go and actually prove a hypothesis.
In this project, an investigation of Linear Pixel Shuffling for subpattern searching was performed using software created for this purpose. Linear Pixel Shuffling is a method to get a random-like permutation of pixels. Because the algorithm is generic, it has been used for displaying images, graphics rendering, image compression, image mor- phology, and digital halftones. The algorithm was compared against Space Domain Correlation, Fre- quency Domain Correlation, and Random Ordering for its quickness in matching a pattern. On average, Linear Pixel Shuffling found a partial match in a third of the time compared to Space Domain, Fre- quency Domain, and Random Ordering.
The title of my paper is Linear Pixel Shuffling and Subpattern Searching. It has been converted from FrameMaker to HTML format. If you would like to see the paper as it was originally formatted, I have also converted it to PostScript.
If that's still not enough, I also gave a slide presentation. As with my paper, there is a PostScript version available.
The software I wrote to run the experiments is freely available. I call the program Sherlock.
Sherlock is an Open Look application written in C using the XView toolkit on a Sun SPARCStation 4/40.
I should warn you that the compressed (using the GNU gzip program) file is roughly 12 MB in size. Not only are you getting the program, but for this low, low price, you are also getting man pages, shell scripts, image files, numerical results, and tons more! Seriously, all three experiments, including the input data, the results, and my paper and slides in Postscript format are packaged in with the program.
Last updated: 02/17/95