An improved SIFT algorithm for image feature-matching

Abstract

SIFT algorithm has strong robustness and stability, which can be applied in many bad conditions to achieve high recognition rate. The 128-dimensional feature descriptor has good independence. However, it also contains redundant information, which makes the calculation amount of following matching increase. An improved SIFT algorithm is proposed to solve the disadvantages of large computation and not meet "real time" of the original one, which reduces the dimension of the feature vector and reduces the mismatching of SIFT algorithm. It has been proved from the experiment that the improved algorithm can significantly increase running speed and guarantee the robustness. Besides, it can meet "real-time" requirements.