Robust Binary Feature Using Intensity Order

Abstract

Binary features have received much attention for memory and computational efficiency as emerging demands of mobile and embedded vision systems. In this context, we present a robust binary feature using intensity order. By analyzing feature regions, we take a simple but effective strategy to detect keypoint. We adopt ordinal description and encode intensity order into binary descriptor with a proper binarization. As a result, our method obtains high repeatability and shows better performance on feature matching with much less storage usage than other conventional features. We evaluate the performance of the proposed binary feature with various experiments, demonstrate its efficiency in terms of storage and computation time, and show the robustness under various geometric and photometric transformations.