[Git Download]


Hypersnakuscules provide an efficient and highly parallel algorithm for evolving spherically symmetric three-dimensional contours. This work extends previous work from EPFL on fast two-dimensional snakes [2]. This formulation is  optimized for 3D images, and includes a strategy for GPU acceleration that takes advantage of Monte-Carlo sampling. Experimental results show that this method provides superior performance for large 2D and 3D cell localization tasks when compared to existing methods on large 3D brain images.


Hypersnakuscules takes as input a 3D RAW image, along with the image size:

>> hypersnakuscules input.raw output.txt --size x y z

The output file output.txt gives the coordinates for each surviving snake, as well as its radius. A full list of command-line options can be listed using:
>> hypersnakuscules --help


[ 1 ] Mahsa Lotfollahi, Sebastian Berisha, Leila Saadatifard, Laura Montier, Jokūbas Žiburkus, David Mayerich, "Three-Dimensional GPU-Accelerated Active Contours for Automated Localization of Cells in Large Images," PLoS ONE (in press)

[ 2 ] Philippe Thevenaz, Michael Unser, "Snakuscules," IEEE Transactions on Image Processing, 17(4):585-593, 14 March 2008