Chapter 7 - Image thinning

Abstract

This chapter defines thinning (or skeletonisation) operations on binary digital images. The specific class of binary line images is defined and modeled in order to adapt thinning operators. This operation results in the skeleton of the image. The chapter also presents models for the skeleton and studies and analyzes different approaches for obtaining it, particularly, a graph-theoretic approach. This context is shown to be well suited for binary line image analysis. The aim is to define invariant image characteristics on which high-level analysis, such as classification and recognition can be performed. The chapter concludes with an introduction to binary line image vectorisation based on the output of the thinning operation. The interpretation of the content of an image relies on the definition of accurate representations on which an analysis process can be applied. For tackling problems, such as image classification and image registration, invariants have to be designed. Basic factors, which can be associated with image components, are not specific enough for uniquely identifying the content of an image under study. This chapter concentrates on the design of a class of models that will allow for such a characterization. These models aim to map the image components onto their skeletons, corresponding to a well-established concept in image analysis. From their skeletons, reliable invariants can be designed.