Optical flow with an intensity-weighted smoothing

Matching algorithms which use image structure such as edges and junctions cannot perform well on unstructured images. For such images, global gradient-based methods may be more appropriate. Some of the assumptions involved in applying discretized conservation equations such as the optical flow equation are discussed and an intensity-weighted method of estimating dense displacement fields is presented, which attempts to avoid some of the problems of the standard methods. The only derivative required is the intensity gradient, and the method includes procedures for automatic parameter evaluation. The algorithm is expected to perform better than conventional algorithms on images without strong texture, as is demonstrated on the test set of images.<
>