Points of interest

anyone have an idea how I might analyse a texture to decide coordinates of ‘points of interest’?

ie if you were to view these points without the original image there would still be a resemblance of some kind

as i learned yesterday from elliot, there is a feature finder in the opencv library. so you could code an EmguCV node to find the features…

What you’r looking for is called “Salience” … if you’r googling for “Salience Using Natural-statistics” you’ll find some papers on the topic. “Salience finding” is a non trivial issue and something different than what you’ll find in the “standard” opencv feature finder which basically just finds points with high contrast such as edges and lines in an image.

If you’t looking for an implementation in opencv I can recommend “Nick’s Machine Perception Toolkit” which you’ll find here http://mplab.ucsd.edu/~nick/NMPT/ … an could eventually be used from c# - that’s something you should ask vux or tonfilm about.

Thanks for the link. Have you used this toolkit before?


used if before but only in c++ land. I’ve written an openframeworks wrapper but can’t find it right now.