What is the relationship between AKAZE descriptor size and the AKAZE feature patch size?

asked 2018-09-04 06:47:09 -0600

stiefel512 gravatar image

I am running the AKAZE feature detector (python) on an extremely large image, and I want to tile the image. In order to actually get the features on the entire image, I want to overlap the tiles with enough pixels to cover half the patch size, but in order to avoid getting redundant points, I want the overlap to be half the patch width.

I see experimentally that the default descriptor size is 61. I also see experimentally that I am not getting any keypoints within 28 pixels of the edge of the image. However, I don't fully understand the details of the descriptor, and so I don't know how to translate that into the default dimensions of the patch.

How can I find the default width of the patch in pixels?

edit retag flag offensive close merge delete

Comments

patch and descriptor size are (somewhat) unrelated.

i found a hint here: https://github.com/opencv/opencv/blob...

Actual patch size is 2*pattern_size*point.scale
berak gravatar imageberak ( 2018-09-04 07:31:04 -0600 )edit
add a comment