FAST SPECTRAL CLUSTERING FOR IMAGE SEGMENTATION

摘要

Standard spectral clustering requires finding the eigenvectors of the affinity matrix of all the data point. The time and space complexity of this operation is a barrier to the scalability of this algorithm. In this paper, we propose a fast spectral clustering algorithm based on Lanczos iteration and fast N-body methods. We apply our algorithm on both synthetic data and the real-world problem of color image segmentation. Our results show that our approach significantly reduce the time and space requirement.