Research on Augmented Reality of Printed Matter Based on Random Ferns and Orientation Gradient with Integral Graph

摘要

Though random ferns is a fast and robust method among the current feature matching algorithms, its way to build local binary feature through comparing single pixel's gray value is not stable with respect to noise, rotation, view variety or illumination changes. Orientation gradient descriptor is more robust than gray level, but the traditional histogram of orientation gradient descriptor is a little complicated and time consuming. This article used an orientation gradient descriptor combined with integral graph to improve the process in random ferns. The experimental results showed that the orientation gradient descriptor with integral graph effectively improved random ferns' matching accuracy while retain the efficiency advantage.