改进SIFT算法实现图像的快速匹配
Fast image match with an improved efficient SIFT algorithm

为解决图像匹配耗时的问题，提出一种改进的图像匹配方案. 在尺度不变特性变换（SIFT)算法的基础上，以特征点邻域灰度值的差熵大小来筛选稳定特征点，减少所需描述及匹配的不稳定特征点的数量，提高算法匹配效率. 同时，改进误匹配去除算法，以大幅提高误匹配去除效率. 实验结果表明，与SIFT及RANSAC相结合的图像匹配方案，或相关的改进方案相比，本方案可最大程度地保存最终匹配的特征点数量，并提高特征点匹配的实时性、 匹配率及正确匹配率.
An improved image matching method has been proposed to accelerate the image matching procedure based on SIFT algorithm. First，an entropy-based approach to filter out the unstable feature points from the candidates is proposed. It can reduce the unstable ones during the feature descriptor computing and matching stages. Second，a fast mismatching-removing method，which highly improve the efficiency of eliminating the small quantity of remaining mismatching points，is proposed. A serial of experiments are conducted which demonstrate that the proposed approaches have a better accuracy，real-time than other improved methods