高斯差分的重要性
我在 beagle board xM 上实现了 sift 的项目，结果还不错......但是对于演示部分，我仍然不明白为什么在 sift 中考虑高斯差异而不是选择 LoG（高斯的拉普拉斯算子） ）。有人可以给我一个答案，这将对我的演示有很大帮助
i did my project implementing sift on beagle board xM and came out of it OK....but for the presentation part,i still don't understand the reason why difference of Gaussian was considered in sift rather than opting LoG(laplacian of gaussian).can somebody please give me an answer which will greatly help my presentation
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
首先，很抱歉耽误了这么长时间。
在 LoG 操作中，首先我们拍摄图像并对其进行一些模糊处理。然后求二阶导数，也称为拉普拉斯算子。它定位用于查找关键点的边缘和角点。但二阶导数对噪声极其敏感。模糊可以消除噪声并稳定二阶导数。
整个问题是计算密集型的。
因此，进行了一个小的调整，找出模糊的图像并找出图像的差异。这与 LoG 大致相同，但计算简单。
希望我以最好的方式解释了这个问题。
First of all, sorry for the long delay.
In a LoG operation, first we take an image and blur it a bit. Then the second order derivaties are found which is also known as laplacian. It locates the edges and corners which are used for finding keypoints. But the second order derivative is extremely sensitive to noise. The blur smoothes it out the noise and stabilizes the second order derivative.
The whole problem is computationally intensive.
Thus a small tweak is made in which blurred images are found out and the difference of the images is found out. This is approximately the same as LoG but is computationally simple.
Hope I explained the question in the best possible way.