How to Exploit Marker Orientation for SIFT Image Matching

Abstract

In computer vision, it is very important to obtain reliable corresponding feature points. However, it is not easy to locate the corresponding feature points exactly considering scaling, lighting, viewpoints, etc. Lots of SIFT methods apply the invariant to image scale and rotation and change in illumination, which is due to the feature vector extracted from the corners or edges of the object. However, The SIFT method cannot find feature points easily when the brightness value of the area is similar since we extract feature points along the edges. In addition, this algorithm determined correspondences by vector difference between local descriptors. In this paper, we propose a marker design and a placement way for improving the performance regarding the detection and matching of feature points. The shape of the markers used in the proposed method is formed in a semicircle in order to detect the predominant direction vector by the standard SIFT algorithm depending on the direction of placement regarding the marker. Also, we propose a key point matched way, which is a combination of each features predominant orientation and nearest neighbor. We examined the features for point detection and correspondence matching. The experimental results demonstrated that the proposed method is more accurate and effective compared to the current method.