Image objects detection with local topological characteristics, obtained by two-dimensional variations

Abstract

I proposed method for producing local topological image characteristics on the base of two–dimensional variations. This method uses variations-based estimates and forms the metrical parameters of the objects of observed scene. I offered new concepts of the object size index and the convexity amplitude indices up and down. I stated algorithm for calculation of these characteristics. Application of these indices renders the new approach for object detection in an image. I demonstrated results of object detection in real gray scale and color images.