The idea of the correlation of parts of one image with parts of the next (in order to find image flow) has been used more often as part of flow estimating algorithms than as a single method in its own right. The method does not perform the data reduction of the feature-based techniques, and is computationally expensive. Also, as with gradient-based methods, correlation-based flow will be the most well conditioned near image features. However, in  and , Burt et. al. use correlation at multiple scales to find optic flow in cases where multiple motion may be occurring. In  correlation and image warping is used whilst moving downwards through scale space to find image flow. See also the references in .
More often, such as in ,  and , correlation is used to aid the matching of image features or to find image motion once features have been determined.