Learning salient structures for the analysis of symmetric patterns

Abstract

Feature-based symmetry detection algorithms have become popular amongst researchers due to their dominance in performance, nevertheless, these approaches are computationally demanding. Also they are reliant on the presence of matched features, therefore they benefit from the abundance of detected keypoints; this implies that a trade-off between performance and computation time must be found. In this paper both issues are addressed, the detection of large sets of keypoints and the computation time for feature-based symmetry detection algorithms. We present an innovative process to learn rotation-invariant salient structures by clustering self-similarities. Keypoints are detected as local maxima in feature-maps computed using the learnt structures. Keypoints are described using BRISK. We consider an axis of symmetry to be a dense cloud of points in a parameter-space, a density-based clustering algorithm is used to find such clouds. Computing times are drastically shortened taking an average of 0.619 s to process an image. Detection results for single and multiple, straight and curved, reflection and glide-reflection symmetries are similar to the current state of the art.