Uncertainty characterization in image based measurements: a preliminary discussion

Measures deriving from digital images are very important in a number of application fields like industrial manufacturer´s inspection. Clearly, to make a profitable quantitative use of such measures, it is indispensable to achieve a complete control on uncertainties that real systems introduce along the chain of steps brought from real world objects to the results of the measurement process. This is a nontrivial task becoming even more difficult in the presence of a large variety of software solutions for achieving measures. In such contests, the availability of a reasonably general mathematical model of uncertainty propagation could be the only way to allow reliable engineering choices among different hypotheses. In the present paper we present such a model and test its application to Canny edge detector both in artificial and real-world images. The validity of the model is finally discussed in respect to a wide amount of experimental results