尺度不变特征转换特征提取优化算法研究

Optimization of scale-invariant feature transform(SIFT) feature extraction

针对尺度不变特征转换(SIFT)算法时间复杂度高的缺点,提出了SIFT特征提取优化算法。分析了SIFT特征提取算法各个计算步骤的时间复杂性。对SIFT特征提取过程进行了优化,包括耗时最多的高斯金字塔的创建和计算特征描述符过程。优化算法降低了特征点提取时间,减少了特征点的重复匹配,同时保证了匹配结果的准确性。最后,实验证明了优化后的算法能有效降低时间复杂度。

Given the limitations imposed by the high time complexity of scale-invariant feature transform （SIFT） al- gorithms, an optimized SIFT feature extraction algorithm has been developed. The time complexity of the calcula- tion steps of the SIFT feature extraction algorithm was first analyzed. Secondly, the Gaussian pyramid creation and feature^point calculation process, which are the most time-consuming processes in a SIFT algorithm, were opti- mized. The optimized algorithm succeeded in reducing the feature point extraction time and reducing the duplicated feature matching, whilst at the same time guaranteeing the accuracy of the matching results. Finally, experiments showed that the optimized algorithm effectively reduced the time complexity.