一种基于图像插值的SIFT算法研究

SIFT特征是图像的局部特征,其对旋转,尺度缩放,亮度变化保持不变性,对视角变化、一定角度的仿射变换、噪声也保持一定程度的稳定性.提出了一种具有尺度不变特征转换的图像插值的SIFT算法.首先对图像进行SIFT特征提取,利用K-D树搜索并计算最近邻特征向量与次最近邻向量间的欧式距离来实现对特征点的匹配.在此基础上加入图像插值法增大采集特征点的范围,从而增加特征点匹配的对数.最后比较3种图像插值方法.实验结果表明：双线性插值法获取的特征点匹配的对数最多,但匹配时间较长.

An algorithm for SIFT based on image interpolation

The SIFT feature is the local features of the image,it remains invariant regardless of rotation, scale, brightness change and it remains stable regardless of the viewing angle changing, the affine transformation and nose. This article proposes a kind of SIFT algorithm with scale-invariant feature transform for image interpolation. First, using SIFT to extract invariant features from images, then using the K-D tree searehes and calculates the nearest feature vector and the second nearest neighbor vector between the Euclidean distance to achieve the matching of feature points. After that, image interpolation method is performed to increase the collect the characteristic range of points in order to increase the number of feature points matching. Finally, it compares three kinds of image interpolation methods. The experimental results show that bilinear interpolation method obtaines the largest number of matching of feature points, but the matching time is too long.