BSIFT: Boosting SIFT Using Principal Component Analysis. M. Fotouhi, S. E. Mirsadeghi, S. Kasaei, K. Faez
This repository contains implementation of the above paper.
Run BSIFT_demo.m to see result of BSIFT on the leuven dataset.
VLFeat -- Vision Lab Features Library https://github.com/vlfeat/vlfeat
TOTAL AND CORRECT MATCHES FOR GRAFFITI SAMPLE
Method | PCA Energy | Equivalent Dimension | I1 -> I2 | I1 -> I3 | I1 -> I4 | I1 -> I5 | I1 -> I6 |
---|---|---|---|---|---|---|---|
BSIFT | 50 % | 8 | 957/1384 | 305/725 | 50/380 | 9/298 | 0/197 |
BSIFT | 60 % | 13 | 986/1473 | 367/882 | 74/475 | 10/366 | 3/293 |
BSIFT | 70 % | 19 | 1005/1533 | 421/995 | 94/572 | 12/441 | 3/350 |
BSIFT | 80 % | 30 | 1010/1554 | 441/1039 | 105/609 | 15/468 | 3/390 |
BSIFT | 90 % | 51 | 1014/1568 | 448/1050 | 111/663 | 18/491 | 2/414 |
SIFT | 100 % | 128 | 997/1505 | 417/1001 | 98/612 | 14/501 | 2/438 |