Post material pulled from machine vision expert Chris Walkers blog.
Image normalization is often used to enhance image contrast. Normalization techniques create a greater difference between the darkest grays (by pushing them towards black) and the lightest grays (by pushing these towards white). The image on the left below is an image before normalization, with a normalized image on the right.
In machine vision, normalization is a useful tool for improving the ability of the system to see certain product features. Increasing image contrast will either improve standout of certain features or refine the vision system’s ability to segment the image.
Equalization is one common type of image normalization that evenly distributes the pixels in an images histogram (spread of shades from black to white). The drawback to equalization is it can also enhance noise in an image, or create image “artifacts” that were not present before.
One example of a vision system where normalization is a benefit is reading dark print on a dark bottle cap (Optical Character Verification OCV). You can see that with a better contrast between areas of the image, the print becomes much more detectable to the camera lens:
You can explore this topic in more technical detail on the blog of machine vision expert Chris Walker.