The term optical flow is used to describe the motion field associated with a sequence of images. For the calculation of two-dimensional optical flow, a variety of methods is used and a good overview can be found in [BFB94]. The central assumption of all methods is that the flow field to be determined can be regarded as being caused by a (locally) constant, moving brightness pattern. This is, of course, a simplified viewpoint, since under general illumination conditions, the projection of a three-dimensional moving (or rotating) object does not yield a constant moving brightness pattern. Another important point is, that depending on the actual pattern it might not always be possible to reconstruct the correct flow field from local information only. As we will see in a later section the latter constraint does not hold for one-dimensional optical flow.
In the following sections we discuss the principles underlying most of the methods for calculating optical flow in one or two dimensions.