What's the size of the patch used to get a feature with HarrisLaplaceFeatureDetector

asked 2019-10-24 04:09:56 -0600

Roysorus gravatar image

updated 2019-10-24 04:12:47 -0600

Hello guys,

I'm trying to use cv::xfeatures2d::HarrisLaplaceFeatureDetector within a grid.

What I do is divide the image into cells and detect in each cell in order to get features all over the image.

When using ORB detector,there is a parameter for patch size, so in each cell I add the patch size (from the top, the bottom, left and right of the cell) in order to be able to have features in all the cell.

However, for HarrisLaplace there is no parameter for patch size. Is there a way to find out the patch size used?

edit retag flag offensive close merge delete

Comments

are there "patches", even ?

https://www.robots.ox.ac.uk/~vgg/rese...

well, maybe: (p.68)

At each level of the representationwe extract the interest points by detecting the local maxima in the 8-neighborhood of a point x

berak gravatar imageberak ( 2019-10-24 04:20:15 -0600 )edit

You seem to be right. So it is a 17 pixels patch?

Roysorus gravatar imageRoysorus ( 2019-10-24 04:30:10 -0600 )edit

can you maybe look, what's really going on here and report back ? we'd probably have an answer to your question, then ;)

berak gravatar imageberak ( 2019-10-24 04:34:20 -0600 )edit

From what I understood, the patch size is : 3*powf(2, oct + (layer/all_layers)). For oct 0 if varies from 3 to 4 pixels. Am I right?

Roysorus gravatar imageRoysorus ( 2019-10-28 04:55:58 -0600 )edit
add a comment