A device to classify surface orientation from polarization images

One way to warrant the accuracy of the measure of the polarization angle phi of light after reflection from some surfaces may be to use a very precisely tuned polarizing filter and control of the fine tuning time evolution. This solution may be expensive and not easy to implement. Another way is to try to have a measurement process as independent as possible of the accuracy of the devices. This is the reason for proposing a method for precisely determining the angle of polarization phi based on self-calibration techniques. The advantage over existing methods is that it is not necessary to exactly know the rotation angle theta of the polarizing filter. From these accurate and robust measures of polarization, an orientation map of the observed objects may be derived to classify homogeneous geometrical regions. This 3D information could be inserted in classical techniques of stereovision to improve the matching process.