This is an implementation of SIFT done entirely in Python with the help of NumPy. A wrapper function, match_template()
, matches a template to an image and displays the result as a demonstration of the SIFT algorithm.
Note: this code relies on OpenCV version 2.4.11.
The wrapper function match_template()
is used to call detect_keypoints()
.
Running from python in terminal:
from siftmatch import match_template
match_template(imagename, templatename, threshold, cutoff)
where imagename
and templatename
are filename strings (e.g., "image.jpg"
), threshold
is the contrast threshold for the sift detector, and cutoff
is the maximum distance between a keypoint descriptor in the image and a keypoint descriptor in the template for the two keypoints to be considered a match. A good value for threshold
is 5
.
Note that if there are too many keypoints, flann.knnSearch()
on line 16
of siftmatch.py
may fail if you don't have enough RAM. Increasing threshold
will reduce the number of keypoints found by SIFT.
Running from python in terminal:
from siftdetector import detect_keypoints
[keypoints, descriptors] = detect_keypoints(imagename, threshold)
where imagename
and threshold
are defined as above, keypoints
is an n
by 4
numpy array that holds the n
keypoints (the first column is the image row coordinate, the second column is the image column coordinate, the third column is the scale, and the fourth column is the orientation as a bin index), and descriptors is an n
by 128
numpy array where each row is the SIFT descriptor for the respective keypoint.