Variation model🔗
The main principle of a variation model is to compare one or more images to an ideal image to find significant differences. With this, you can, for example, identify incorrectly manufactured objects by comparing them to correctly manufactured objects. The ideal image is often obtained by a training using several reference images. Besides the ideal image, the training derives information about the allowed gray value variation for each point of the image. This information is stored in a so-called variation image. Both images are used to create a variation model, to which other images can be compared.
The advantage of the variation model is that images can be directly compared by their gray values and the comparison is spatially weighted by the variation image. Below, see a example for a variation model that is used to inspect logo prints on a pen clip.