A Practical Introduction to Computer Vision with OpenCV

6 Edges

The segmentation of an image is the process of dividing the image into pieces such that different objects or parts of objects are separated. This is an essential processing step in image understanding (where the goal is to understand the content of the images). There are two main ways of addressing the segmentation of images:

1. Edge processing, where we identify the discontinuities (the edges) in images.
2. Region processing, where we look for homogeneous regions (or sections) of the images. Binary vision is a very simple example of region processing and much more complex approaches to this problem exist (although this is not a topic considered in this text).

These representations can or should be complementary, with the edges delineating the homogeneous regions. Unfortunately it is very difficult to decide where the edges are (and hence to get a unique edge image) and it is equally hard to make a unique interpretation of an image in terms of regions. There are many techniques in both edge-based vision and region-based vision, all of which have the goal of determining the best representation of the scene in terms of edges or regions. For most images there is no absolute correct answer as the answer is typically subjective (it will depend on the observer and will depend upon the purpose of the segmentation).

This chapter looks at basic edge detection techniques for determining which pixels should be considered to be edge pixels (see Section 6.1), how these edges points can ...