Computer Vision News - July 2016

One can notice that the image is a bit more blurry, but the edges and the image structure are nicely preserved. In term of computation, we now have 15% of the original pixels. Computing now the affinity matrix takes less than one second (it was 3 minutes in the original image). Spectral-Clustering with four clusters will yield the following results. Computer Vision News Tool 23 With some simple morphological operation one could easily extract the desired four image regions: Pony body, Pony hear, the Tuba and background. Where to find the implementation The algorithm itself is not hard to implement. But why to reinvent the wheel if you can find it online on the web ? The SLIC SuperPixels is implemented in almost every programing language and image processing package. To save you time we have gathered for you the information about where to find it: Note that although the SLIC algorithm is the most popular SuperPixel algorithm, it is not the only one out there and many flavors of SuperPixel algorithm do exists. For a competition between 7 state-of-the-art SuperPixel algorithms, with a total of 9 implementations used for evaluation, we invite our readers to learn more here . Programing language / package Link + details C++ / openCV API Installation instruction Python Example C http://ivrl.epfl.ch/research/superpixels Matlab Here Tool