www.pudn.com > PCA-SIFT.rar > README

PCA-SIFT
by Yan Ke

This code implements the PCA-SIFT algorithm.  Interest point detection
code must be obtained separately.  Any scale and orientation invariant
interest point detector will work.  Interest point representation is done
through PCA gradient vectors.

Given an image in PGM format, outputs keypoints in ascii text.
The format is (mostly) compatible with David Lowe's keypoint format.
Use mod_lowe_demoV2.tar.gz to detect keypoints and match results.

Usage:

./keypoints < image1.pgm > image1.lkeys
./keypoints < image2.pgm > image2.lkeys

./recalckeys gpcavects.txt image1.pgm image1.lkeys image1.pkeys
./recalckeys gpcavects.txt image2.pgm image2.lkeys image2.pkeys

./match -im1 image1.pgm -k1 image1.pkeys -im2 image2.pgm -k2 image2.pkeys > out.pgm

"keypoints" and "match" can be found in mod_lowe_demoV2.tar.gz

For more information, visit:

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