We present a novel geometry-based method for computing asymmetric tunnels and voids in proteins, approximating their real shape with selected precision. Our method combines ideas from Voronoi and grid based approaches for protein analysis. We represent tunnels in protein using voxel data grid which allows us to store their shape more accurately. Our algorithm employs a tunnel skeleton computed using Voronoi diagram. The skeleton allows us to perform grid computation in a bounded space, with lower time and memory demands, and easily identify and measure individual tunnels.