Classical stereo matching techniques use features representing objects
in both images and compute the minimum difference attribute values. N
o knowledge of the environment is taken into account. This paper propo
ses an image understanding stereo matching method using a supervised n
etworks: the perceptron.