Scale Invariant Feature Transform (SIFT) is one of the most widely recognized feature detection algorithms. It was proposed by David Lowe in 2004.
Note
Link to the paper: http://www.cs.ubc.ca/~lowe/papers/ijcv04.pdf
Some of the properties of SIFT are as follows:
It is invariant to scaling and rotation changes in objects
It is also partially invariant to 3D viewpoint and illumination changes
A large number of keypoints (features) can be extracted from a single image
SIFT follows a strategy of matching robust local features. It is divided into four parts:
Scale-space extrema detection
Keypoint localization
Orientation assignment
Keypoint descriptor
