Preprocessing

Abstract

In the previous chapter we have concerned ourselves with the basic theory of shape detection using parametric transformation. We have arrived at the point where we can confidently cope with straight line detection using computer generated binary data. In order to consolidate our position we will deal only with straight lines for the time being. This is because our next steps will bring us into head-on collision with the real world. This place is at one time noisy and packed full of confusing information. To begin with, a real image will not be conveniently binary. It will present itself as a two dimensional array of pixels or feature points each containing a discrete intensity value. The intensity values may vary over a relatively large range, for example [1, 256]. Some of the feature points will truly carry information about the real world. Others may have been corrupted in some unexpected way in the physical process of image acquisition. They are what we call noise.