Ask Your Question

What is the best solution for rotation invariant detector?

asked 2013-02-14 01:22:15 -0600

lizris gravatar image

I'd like to create an object detector based on cascade classifier, the only problem is that LBP and Haar features are not rotation invariant. There are many rotation invariant detectors, for example, iPhone recognizes faces in real time in any orientation, so I wonder how do they achieve this? The first thing that comes to my mind is to rotate the training sample at different angles, but I doubt that the resulting classifier would have good quality, moreover, the object could have stretched proportions.

edit retag flag offensive close merge delete

2 answers

Sort by ยป oldest newest most voted

answered 2013-02-14 04:17:11 -0600

Guanta gravatar image

LBP can be made rotational invariant. Compute all cyclic shifts and take the one which gives you the minimum.

edit flag offensive delete link more


Could you provide some more information about how to achieve this in OpenCV?

lizris gravatar imagelizris ( 2013-02-14 08:00:52 -0600 )edit

You can shift a value using '<<' odr '>>' oerator. See e.g. , see also for a tutorial on lbp which also explains the rotational invariance by this step.

Guanta gravatar imageGuanta ( 2013-02-14 11:13:03 -0600 )edit

answered 2013-02-14 03:24:27 -0600

updated 2013-02-14 03:26:07 -0600

One solution for rotation invariant face detection is below steps:

1- Use skin detection for estimate face blobs . 

2- Fit ellipse on every blob for recognition dimension of blob .

3- Clear extra blobs.

4- Compute the average of ellipses angle. 

5- Rotate input image by above angle.
edit flag offensive delete link more


I need to detect logos, not faces, moreover, input images could be in grayscale.

lizris gravatar imagelizris ( 2013-02-14 08:00:01 -0600 )edit

Hi Lizris, I know this is kind of an old question. Did you find a workable answer to this?

  • Karthik
Samjakar gravatar imageSamjakar ( 2016-02-10 04:13:05 -0600 )edit

Question Tools

1 follower


Asked: 2013-02-14 01:22:15 -0600

Seen: 3,036 times

Last updated: Feb 14 '13