Hypergraph-Based Image Processing

摘要

Learning and processing of signals over hypergraph models have gained substantial traction owing to the ability of hypergraphs in characterizing multilateral interactions. In this work, we explore hypergraph spectral analysis and provide alternative definitions of frequency domain operations that are practically useful in image processing. We analyze hypergraph spectral properties and present several application examples, including compression, edge detection and segmentation. Successful experiment results demonstrate the effectiveness and the future prospect of the proposed hypergraph frequency operations in image processing.