Introduction to computerized image analysis


With classification we mean that we would like to say that this specific pixel p belongs to the class forest, this belongs to open field, and this belongs to the class road. We do this by calculating various features. Those are used to put up a model for a typical forest pixel, a typical open field pixel and a typical road pixel. A pixel in the image is then classified as belonging to the class which it most probable that it actually belongs to from the feature it has.

Below to the right, we see an aerial image over a forest. We can easily identify that there is a region with forest, through it a road is passing, and that there is also a region which is open field. With the help of a classification method we can obtain the image to the right, where pixels that the algorithm has classified as forest are marked in red, open field in yellow and road in blue.

skog klassad skog
Aerial image of forest Classification of the image

In the classification step it is usual to use different statical methods to decide which class a pixel or object belongs to.