The Comparison and Analysis of Scale-Invariant Descriptors based on the SIFT

Based on the feature matching theory about SIFT (Scale-Invariant Feature Transform) keypoints, the concentric circle structure and the color feature vector of scale-invariant descriptor are proposed in this paper. In the concentric circle structure, the radiuses of the concentric circles are proportional to the scale factor, which can achieve the scale invariance. To achieve the rotation invariance, the coordinates of descriptor are also rotated in relation to the point’s orientation. Compared with the square structure of SIFT descriptor, the concentric circle structure not only has simpler computation, but also is more robust to image rotation. The color feature vector chooses the mean values of different color components R, G, B in each subregion of descriptor as the vector’s elements. Compared with the gray feature vector of SIFT descriptor, the color feature vector fully utilizes the image’s color information, having stronger rotation invariance, and obviously decreasing the vector’s dimension, with less computation. After the theory analyses, the experimental results have certified their validity, too. Key-Words: computer vision; feature matching; scale-invariant keypoint; scale-invariant descriptor; concentric circle structure; color feature vector;