My results (Bug #1044)

Added by Julius Adorf almost 14 years ago. Updated about 13 years ago.

Status:Cancelled Start date:
Priority:Blocker Due date:
Assignee:Maria Dimashova % Done:

0%
Category:features2d
Target version:-
Affected version: Operating System:
Difficulty: HW Platform:
Pull request:

Description


You can reproduce this bug using the small code snippet attached, the example image and the mask specifying the region of interest.

Filtering the keypoints after detection is hurting performance. The example image is 1280x1024 and the ROI includes roughly 33.000 pixels (which is only 1/40 of the whole image). I don't know too much about the computational complexity of SIFT, but it will be at least linear in the number of pixels of the supplied image. Experiments show an slowdown way more than expected after switching from FAST to SIFT.

siftmask.cpp - run to reproduce bug (1.2 kB) Julius Adorf, 2011-05-04 04:09 pm

image_00000.png - an example image (1.7 MB) Julius Adorf, 2011-05-04 04:10 pm

image_00000.png.mask.png - and a corresponding mask (161.6 kB) Julius Adorf, 2011-05-04 04:10 pm

Associated revisions

Revision 66c116ec
Added by Maria Dimashova almost 14 years ago

fixed #1044

History

#1
Updated by Maria Dimashova almost 14 years ago

Fixed in r5024.
Thanks for the bug report.

  • Status changed from Open to Done
  • (deleted custom field) set to fixed

#2
Updated by Julius Adorf almost 14 years ago

Thanks. Manually tested (see attachments). Works for me.

#3
Updated by Bruno Nascimento over 13 years ago

Sorry i've installed version 2.3.1 of OpenCV and this bug is still active.
The mask parameter is being ignored and keypoints are found outside the region of interest.

  • Status changed from Done to Cancelled
  • (deleted custom field) deleted (fixed)

#4
Updated by Kazuki Matsuda about 13 years ago

I run your problem. And get an attached image. I don't identify problem.

Also available in: Atom PDF