Non-maximum suppression in OpenCV

Question

I applied a harris corner detection using openCV which gave me a response map for the potential corners. The documentation states that corners can be found as the local maxima of this response map - does anyone know how this can be done?

Thank you.

Solution

There is a working implementation with full coding for both Harris and the Shi-Tomasi corner detectors here. The 4th parameter in the corner detector function is a CvPoint2D32f* structure. The function returns the calculated local maxima i.e. the corners into this structure array. Simply looping through the array and receiving each point as a CvPoint will get you the corner positions.

(all this is there in the code)

Hope this helps.