如何可视化 SIFT 关键点
我有很多 SIFT 关键点（一些 128 维向量），给定 SIFT 关键点，我想获得可视化图片（如 两张小照片右侧，我该如何做到这一点？我需要几行 MATLAB 代码来实现这一点，有人可以帮我一个忙吗？ 非常感谢。
I have got a lot of SIFT keypionts(some 128-dimension vectors), given a SIFT keypoints, I want to get the visualization picture(as the two tiny pics on the right side, how can I do that? I need servral lines of MATLAB code to achieve this, dose anybody can do me a favor?
many thanks.
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（2）
右侧的图像看起来可能是导致关键点选择的高斯差分图像。
如果您有权访问 SIFT 算法的内部工作原理，则可以从那里获取它们。当检测到稳定的最大值时，保留导致其检测的高斯差分图像。
如果您无法访问 DoG 图像，则工作量会增加一些。您必须根据关键点信息自行计算高斯图像的差异。如果关键点的比例存储在 sigma 中并且您有 vlfeat (www.vlfeat.org)，您应该能够（大约）如下计算它：
并且您的小图像将是一个正方形您感兴趣的关键点位置周围的区域。
我不确定这是否真的是那些图像，但这是我最好的猜测。
The images on the right look like they might be the difference-of-Gaussians image that led to the selection of the keypoints.
If you have access to the inner workings of your SIFT algorithm, you can just take them from there. When a stable maxima is detected, keep the Difference-of-Gaussian image that led to its detection.
If you don't have access to the DoG images, it will be a little more work. You'll have to compute the Difference of Gaussian image yourself given the keypoint information. If the scale of your keypoint is stored in
sigmaand you have vlfeat (www.vlfeat.org), you should be able to compute it (approximately) as follows:and your little image will be a square region around the location of your keypoint of interest.
I'm not sure if this is actually what those images are, but it's my best guess.
要可视化两个图像中相应的 SIFT 点，您可以按照 David Lowe 的 SIFT 演示 在 match.m 中
检查之后的部分
希望这有帮助。
For visualization of corresponding SIFT points in two images you can do as done in David Lowe's SIFT demo in match.m
Check the portion after
Hope this helps.