OPENCV Study notes (vi) SURF study notes

Original articles, reproduced please specify the source: http://blog.csdn.net/crzy_sparrow/article/details/7392345

I am very vegetables, there must be a lot of mistakes, I hope you are not wrong.

After looking at the Harris Corner Point Detection, began to study surf corner detection, found quite complex, a moment also only understand the approximate, the understanding of things summed up, so that the next time to further study.

Surf Corner detection algorithm is a kind of improvement to sift, the main body is now more efficient in speed. The main difference between it and sift is that the method of constructing multi-scale space of image is different.

In the field of computational vision, the scale space is symbolically expressed as an image pyramid, and a Gaussian pyramid is generally used for descending samples. , the input image function repeats the kernel convolution of the Gaussian function and repeats the second sampling, this method is mainly used for the implementation of the SIFT algorithm, but each layer of the image depends on the original image (the current scale may and the original image scale difference is very large, at this time again that the original picture convolution is a bit two), And the image needs to be reset size , so the calculation method is large. Surf algorithm to operate the integral image, convolution is only related to the previous image, the method of reducing the sample is to apply for the size of the image kernel , which is also the SIFT algorithm and surf algorithm in the use of pyramid principle is different. The surf algorithm agrees that the multi-layer image of scale space is processed at the same time, and the image is not sampled twice, thus improving the performance of the algorithm.

Other aspects of the difference, to my rookie level of understanding, not much difference.

Attached to two SIFT related blog posts, or more thorough, two articles combined to see the basic clarity of its principles.

Two of the things are a little bit in and out, just a comparison to see or to find out the problem, of course, the most easy to find out the problem of the paper.

Sift Brief introduction: http://www.360doc.com/content/11/1230/23/3054335_176200661.shtml

Sift algorithm experience: http://www.360doc.com/content/11/1207/17/3054335_170430459.shtml

Other blog posts:

Deep analysis of Sift/surf algorithm--on the subtlety and insufficiency of SIFT: http://hi.baidu.com/xiaoduo170/blog/item/a22bcc1c2349708286d6b636.html

Sift/surf Series: http://www.yongblog.com/archives/tag/surf%E7%AE%97%E6%B3%95

Surf thesis and source code (the thesis is very specific, code C + + implementation):

http://download.csdn.net/detail/crzy_sparrow/4171374

OPENCV Small Test Surf algorithm:

#include "opencv2/opencv.hpp" int main () {Cv::mat  image, Image1 = Cv::imread ("test.jpg");    Gray-scale Transformation    cv::cvtcolor (Image1,image,cv_bgr2gray);   std::vector<cv::keypoint> keypoints;   CV:: Surffeaturedetector Surf (2500);   Surf.detect (image,keypoints);   CV::d rawkeypoints (Image,keypoints,image,cv::scalar::all (255), CV::D rawmatchesflags::D raw_rich_keypoints);    Cv::namedwindow ("Surf");    Cv::imshow ("Surf", image);    Cv::waitkey (0);    return 0;}

Test Results:

The radius of the marker ring is related to the scale of the feature point, which is the direction of the feature point.

tutorials, source code and application software for SIFT algorithms
1. Founder of Ubc:david LOWE---Sift algorithm, two giant classics
http://www.cs.ubc.ca/~lowe/

2, Cmu:yanke---pcasift, summary of the SIFT aspects of the article
http://www.andrew.cmu.edu/user/yke/

3, Ubc:m.brown---Sift algorithm for the classic application of image stitching Autopano-sift, including a siftlib library
Http://www.cs.ubc.ca/~mbrown/autostitch/autostitch.html
Http://www.cs.ubc.ca/~mbrown/panorama/panorama.html

4, Toronto:jepson---Matlab SIFT tutorial, super Super Super Classic ~

http://www.cs.toronto.edu/~jepson/csc2503/

5, Ucla:vedaldi---The University of California, a doctoral student compiled matlab SIFT tutorial
http://www.cs.ucla.edu/~vedaldi/

6.http://en.wikipedia.org/wiki/scale-inva ... _transform

7. Computer vision classification of Daniel's finishing

Http://www.cs.ubc.ca/~lowe/vision.html

8. http://note.sonots.com/SciSoftware/SIFT.html

9. Reference to the Ransac method for computing the transformation matrix

Http://web.engr.oregonstate.edu/~hess/index.html

10. Affine invariant feature point detection, which refers to the method of performance evaluation

http://www.robots.ox.ac.uk/~vgg/research/affine/

11. A Japanese, pretty bull.

http://note.sonots.com/

Pca-sift.

http://www.cs.cmu.edu/~yke/pcasift/

OpenCV Sift

Http://web.engr.oregonstate.edu/~hess/index.html

Matlab Sift

Http://www.vlfeat.org/~vedaldi/code/sift.html

Http://www.vlfeat.org/overview/sift.html

Improve scale invariant Feature Transform (SIFT) Stanford

Http://robots.stanford.edu/cs223b04/project9.html

Known implementations of SIFT MIT

Http://people.csail.mit.edu/albert/ladypack/wiki/index.php/Known_implementations_of_SIFT

OPENCV Study notes (vi) SURF study notes