SBRISK: speed-up binary robust invariant scalable keypoints

Abstract

Keypoint generation, including detection, description and matching is the basis of a broad range of applications. A more efficient and effective keypoint generation method is always of interest. In this paper, we propose the speed-up BRISK (SBRISK), a variant of the binary robust invariant scalable keypoint (BRISK). SBRISK not only inherits the high speed of BRISK in the keypoint detection, but also adopts a nearly circular symmetric constellation to describe the pattern of keypoint. To adapt to the characteristic orientation of keypoint, SBRISK shifts the binary vector rather than rotating the image pattern or constellation like many other descriptors have done. It abandons interpolation to get intensity at sub-pixel position, since the constellation does not strictly restrict to circular symmetric. Different from BRISK, SBRISK classifies keypoints into bright patterns and dark patterns. Comparison is conducted only within the same class. Meanwhile, a special refinement scheme is imposed upon the initial matching results to improve the match precision. Experiments show that SBRISK has a faster and better performance than BRISK with less memory consumption.