An efficient euler number computing algorithm

The Euler number of a binary image is a very important topological property for image processing pattern recognition, and computer vision. This paper proposes an efficient run-based algorithm for Euler number computing. For calculating 8-neighbor runs, unlike the conventional run-based algorithm, which needs to record start points and end points of all runs, our algorithm only needs to record end points of runs in odd rows. By our method, the number of pixel accesses required for computing Euler number will decrease. Experimental results on various noise images demonstrated that our algorithm outperforms conventional Euler number computing algorithms.