Shape centered interest points for feature grouping

Abstract

Image encoding using interest points is a common technique in computer vision. In this paper we present a scale and rotation invariant shape centered interest point (SCIP) detector. By means of detecting singularities in Gradient Vector Flow (GVF) fields we find points of high symmetry in the image. Due to the nature of the underlying GVF field we can employ our features to group together edge-based interest points such as SIFTs. This feature grouping provides a strong descriptor for SCIPs and can help to encode valuable information about the image for computer vision tasks. We demonstrate the main properties of our features such as scale and rotation invariance and further robustness against noise and clutter in a series of experiments. We show that the information they encode is to a certain degree complementary to SIFT. Furthermore, we evaluate them in an edge map reconstruction task to assess the amount of image information they encode. Finally, we show the power of feature grouping with our framework in a multi-category classification task on natural images from the StreetScenes database.

Robust feature matching for aerial visual odometry

“…The grouping process follows a strategy inspired from [21]. The vectors in the GVF field are followed starting from ORB features.…”

“…Medial keypoints extraction is based on Gradient Vector Flow (GVF) field introduced by Xu and Prince [20] and implemented in [21]. The GVF at a point P (x, y) in the image is the vector field V (P ) = [u(P ), v(P )] that minimizes the energy function given by…”