MARFE instabilities may reduce confinement leading to harmful disruptions. They cause a significant increase in impurity radiation and therefore they leave a clear signature in the video data. This information can be exploited for automatic identification and tracking. A MARFE classifier, based on the phase congruency (PhC) theory, has been developed and adjusted to extract the structural information in the images of JET cameras. This approach has the advantage to use a dimensionless quantity and to provide information that is invariant to image illumination, contrast and magnification. The method is tested on JET experimental data and has proved to provide a good prediction rate.