My image processing class comes to an end, and with it I had to implement an algorithm of my choice. I chose the Simple Linear Iterative Clustering (SLIC) Algorithm because the whitepaper looked really cool, and I'm quite glad I did because I had a lot of fun programming it. This algorithm is used to perform image segmentation, which is useful in many fields like medical imaging or computer vision for example. Furthermore, the segmentation is done using superpixels: geographically compact regions with similar colors. In fact, all operations are done in the LAB color space to benefit from its perceptually uniform property, really important to determine how different two colors are. This novel approach was particularly interesting when it came out in 2010 because it was both faster and easier to use than most contemporary State-of-the-Art approaches using superpixels. You will find below a few results I obtained after running the algorithm:

Original church picture Original
Segmentation using 256 superpixels 256 superpixels
Segmentation using 512 superpixels 512 superpixels
Original fish picture Original
Segmentation using 512 superpixels 512 superpixels
Segmentation using 1024 superpixels 1024 superpixels

If you are interested in learning more, I wrote a more technical report (in french) describing the algorithm in more detail, as well as giving more examples. The code source for this project is free of any rights and is available on my github.