Machine Vision, 3rd Edition

CHAPTER 5 Edge Detection

5.1 Introduction

In Chapter 4 segmentation was tackled by the general approach of finding regions of uniformity in images—on the basis that the areas found in this way would have a fair likelihood of coinciding with the surfaces and facets of objects. The most computationally efficient means of following this approach was that of thresholding, but this turns out for real images to be failure-prone or else quite difficult to implement satisfactorily. Indeed, making it work well seems to require a multiresolution or hierarchical approach, coupled with sensitive measures for obtaining suitable local thresholds. Such measures have to take account of local intensity gradients as well as pixel intensities, and the possibility ...