K-MEANS COLOR CLUSTERING ################## What is K-Means? ################# K-means is an algorithm that is used in many different problem domains. It works by creating centroids, or clusters, that hold specific values - whether RGB color values or an arbitrary size threshold - and then assigning data points to clusters based some distance measurement. For instance, in this program, the user selects four clusters by clicking on points in the window. The RGB and XY values of the selected point become the cluster values. Then, by iterating through every pixel in the image, we can check which pixels belong to which cluters using a Euclidean distance metric. This process is run until all pixels have been normalized into a cluster, resulting in a posterizing type effect. Although this a rudimentary application of K-means, it can still be used to develop powerful and interesting programs. My next project is focused on using K-means to transform photographs into sprite images reminscient of vintage video games. In that case, the centroids will be composed of the color values for specific templates (Commodore 64, Atari 2600, NES, etc). ####################### How to use this program ####################### NOTE: YOU MUST HAVE JAVA AND PROCESSING INSTALLED TO RUN THIS PROGRAM This program only works for Mac and PC currently. If you really want to run it on a GNU/Linux system, you need to install the GSVideo library and import it in the processing code (import GSVideo.*) and then switch the Capture() statements to the appropiate GSVideo funtion. Using the program is simple: if you have a webcam installed, simply open kmeans.pde in Processing and hit the "run" button. Then select four points and watch the algorithm go to work! 'r' and 'y' buttons on the keyboard swtich between RGB and YUV color spaces respectively. To reset the image and select new clusters, press the 'n' key. If you DO NOT HAVE A WEBCAM, you can still use the program but only a static image file. Open kmeans.pde in your editor of choice, comment out lines 13, 14, 47, 58, and 68-70. Uncomment line 46 and change "test.jpg" to the name of whatever file you want to use. Store this file in a folder called "data" in your kmeans directory. Run the program as usual.