文件名称:image-matching-
介绍说明－－下载内容均来自于网络，请自行研究使用
针对
１２８
维
ＳＩＦＴ
特
征向量，采用距离匹配和余弦相似度匹配相结合的测度方法，利用特征点方向一致性进一步降低误匹配率
．
实验结
果表明：改进算法对图像的缩放、旋转、光照、噪声和小尺度的视角变换均有较好的匹配效果
．
与原算法相比，在保
证匹配点数和匹配时间的基础上，改进算法对旋转、缩放、噪声模糊和光照变换的误匹配率平均降低
１０％～２０％
，
对于小尺度的视角变换，误匹配率平均降低
５％．
-For 128-dimensional SIFT feature vectors, using distance matching and cosine similarity measure combining matching method using feature points consistent direction to further reduce false match rate. The experimental results showed that: improved algorithm for image scaling, rotation, light, noise, and small-scale perspective transformation have better matching results. Compared with the original algorithm, while ensuring match points and matching time, based on the improved algorithm for rotation, scaling, noise, blur and light transform the average rate mismatch is reduced by 10 to 20 , for small-scale perspective transformation, mismatch average rate reduction of 5 .
１２８
维
ＳＩＦＴ
特
征向量，采用距离匹配和余弦相似度匹配相结合的测度方法，利用特征点方向一致性进一步降低误匹配率
．
实验结
果表明：改进算法对图像的缩放、旋转、光照、噪声和小尺度的视角变换均有较好的匹配效果
．
与原算法相比，在保
证匹配点数和匹配时间的基础上，改进算法对旋转、缩放、噪声模糊和光照变换的误匹配率平均降低
１０％～２０％
，
对于小尺度的视角变换，误匹配率平均降低
５％．
-For 128-dimensional SIFT feature vectors, using distance matching and cosine similarity measure combining matching method using feature points consistent direction to further reduce false match rate. The experimental results showed that: improved algorithm for image scaling, rotation, light, noise, and small-scale perspective transformation have better matching results. Compared with the original algorithm, while ensuring match points and matching time, based on the improved algorithm for rotation, scaling, noise, blur and light transform the average rate mismatch is reduced by 10 to 20 , for small-scale perspective transformation, mismatch average rate reduction of 5 .
(系统自动生成,下载前可以参看下载内容)
下载文件列表
基于SIFT算子的图像匹配算法研究_白廷柱.caj