一种基于小波变换的快速SIFT图像拼接算法

SIFT特征匹配算法是当前图像拼接研究领域的热点。为改进SIFT算法在图像拼接过程中特征点提取环节计算量大、耗时较长的缺点。本文在研究SIFT算法的基础上，提出一种改进的图像拼接算法。此算法先通过小波变换对图像进行预处理，减少在SIFT尺度不变特征提取过程中的数据计算量，以提高图像拼接速度。对两幅待拼接图像，分别采用经典SIFT算法和本文提出的改进算法做仿真实验，当匹配点控制系数a=0．5，经典SIFT算法在确定特征点耗时1．192891S，改进后的算法仅用了0．856712s。仿真对比试验的结果验证了该算法的有效性，同时表明该算法能够在保持图像拼接效果的基础上，提高图像拼接算法的速度和准确性。

A Fast SIFT Image Mosaic Algorithm Based on Wavelet Transformation

Recently, SIFT feature matching algorithm is becoming the focus of the image mosaic. In traditional SIFT algorithm, the mosaic procedure is computationally intensive and time-comsuming. For solving this problem, an improved SIFT algorithm is pres- ented in this paper. The proposed algorithm combing wavelet transform into SIFT to simplify the scale-invariant feature extraction process, and speed up the image mosaic. The tests of two images mosaicing with the classical SIFT algorithm and the algorithm proposed in this paper have done respectively. With the control parameter a=0. 5, the classical SIFT algorithm took 1. 192 891 seconds to extract the feature points, while the improved algorithm took 0. 856 712 seconds. The contrast simulation tests demon- strate the validity of the proposed algorithm, and show that the speed and accuracy of the proposed algorithm is improved, at the same time the effect of the image mosaic is maintained.