Sift,the scale invariant Feature Transform, the invariant feature transform is an effective method to detect the uniqueness of images, to translate, rotate, scale and even affine transformations (such as taking pictures from different angles) to maintain invariant image local features. It can be easily applied to image matching applications, such as target detection and recognition, or the calculation of geometric transformations between images.
The algorithm is made by David. Lowe was published in 2004 and filed for patent in 2007.
The network already has several versions of the implementation
Please view SIFT's founder's code: David Lowe's SIFT codes:
Http://www.cs.ubc.ca/~lowe/keypoints
But it seems easier to use the SIFT library that Rob Hess maintains:
http://blogs.oregonstate.edu/hess/code/sift/
But when using the Rob code to find an error in its non-critical code, and his code is only suitable for verification, I made some small changes to make it more convenient to use, you can then command line to enter the image, and save the result as a picture and text. The file name is hard-coded.
: http://download.csdn.net/source/2962313
Let's look at the effect:
Feature diagram:
Feature Matching
But the SIFT feature application is also limited, in my checkerboard image is very obvious, please carefully look at the feature point matching
Related to a very good site
Oxford Visual Geometry Research Group
Visual Geometry Group at Oxford
Http://www.robots.ox.ac.uk:5000/~vgg/research/affine/index.html
http://blog.csdn.net/onezeros/article/details/6117704
Pattern matching SIFT---Sift image feature extraction and matching algorithm code