How the descriptor pattern of FREAK is defined?

asked 2012-12-11 18:10:12 -0600

zurwolf
1 ●4

updated 2012-12-15 05:38:28 -0600

I've been looking for some documentation of how the FREAK descriptor pattern is build, but neither the original paper nor the OpenCV implementation have information more than the implementation itself.

My question is, someone knows or have information about how is this defined?

edit retag flag offensive close merge delete

add a comment

1 answer

Sort by » oldest newest most voted

answered 2012-12-15 06:01:30 -0600

zurwolf
1 ●4

I found my response.

The pattern in the OpenCV implementation is defined to obtain a layout like the one described in the paper, with 43 Gaussians, 6 for every size except for the central one, and a response similar to the retina.

However, there isn't a right way build such a pattern as long as the retinas response and the ganglion cells distribution are reproduced.