Abstract.
Calculating a reliable similarity measure between pixel fea- tures is essential for many computer vision and image processing applica- tions. We propose a similarity measure (affinity) between pixel features, which depends on the feature space histogram of the image. We use the observation that clusters in the feature space histogram are typi- cally smooth and roughly convex. Given two feature points we adjust their similarity according to the bottleneck in the histogram values on the straight line between them. We call our new similarities Bottleneck Affinities. These measures are computed e?ciently, we demonstrate su- perior segmentation results compared to the use of the Euclidean metric.