image matching using sift algorithm with global context and local shaped based descriptor

Question

Can anyone explain me how the sift algorithm is working using numerical values..I am searching over Internet but the thing I am getting is only input and output of the sift algorithm but the thing I want that what actually is going on and how,they are calculating keypoints using,global descriptors and how the match between two image is done.

Answer 1

I think you can start by reading the scientific paper.

http://www.cs.berkeley.edu/~malik/cs294/lowe-ijcv04.pdf