The environment that surrounds us is very complex. Understanding and interpreting it is a very hard task. This paper proposes an approach allowing simple form recognition with a camera by using a probabilistic approach called Bayesian Programming. The main goal is to recognize several type of elemental features composing an image, such as local orientation of a contour, or corners. The Bayesian Program for feature recognition is presented and the learning stage explained. One approach has been validated through experiments.