Image Segmentation by Bilayer Superpixel Grouping

Michael Ying Yang
2013 2013 2nd IAPR Asian Conference on Pattern Recognition  
The task of image segmentation is to group image pixels into visually meaningful objects. It has long been a challenging problem in computer vision and image processing. In this paper we address the segmentation as a superpixel grouping problem. We propose a novel graphbased segmentation framework which is able to integrate different cues from bilayer superpixels simultaneously. The key idea is that segmentation is formulated as grouping a subset of superpixels that partitions a bilayer graph
more » ... s a bilayer graph over superpixels, with graph edges encoding superpixel similarity. We first construct a bipartite graph incorporating superpixel cue and long-range cue. Furthermore, mid-range cue is also incorporated in a hybrid graph model. Segmentation is solved by spectral clustering. Our approach is fully automatic, bottomup, and unsupervised. We evaluate our proposed framework by comparing it to other generic segmentation approaches on the state-of-the-art benchmark database.
doi:10.1109/acpr.2013.62 dblp:conf/acpr/Yang13 fatcat:344hex4ifvc7tahhc752wke6mu