Consensus-Based Image Segmentation via Topological Persistence

Qian Ge, Edgar Lobaton; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2016, pp. 95-102

Image segmentation is one of the most important low-level operation in image processing and computer vision. It is unlikely for a single algorithm with a fixed set of parameters to segment various images successfully due to variations between images. However, it can be observed that the desired boundaries are often detected more consistently than other ones in the output of state-of-the-art algorithms. In this paper, we propose a new approach to capture the consensus information from a segmentation set obtained by different algorithms. The present probability of a segment curve is estimated based on our probabilistic segmentation model. A connectivity probability map is constructed and persistent segments are extracted by applying topological persistence to the map. Finally, a robust segmentation is obtained with the detection of certain segment curves guaranteed. The experiments demonstrate our approach is able to consistently capture the curves present within the segmentation set.