Morphological Boundary Pixel Classification

Abstract

This paper presents the method which allow classifying pixels of the internal boundary of objects on the binary image. The application of morphological tools allows grouping internal boundary pixels into four spatial classes: core boundary, isolated regions, branches and corridors. The classification of boundary pixels provides us with a tool for shape description both locally and globally. This description is used to characterize the pixels themselves as well as to describe the image objects. In the latter case the coefficients based on the quantity of pixels of different classes are calculated. These coefficients may be used as features for pattern recognition.