Motion estimation based on the direction of intensity gradient

Abstract

The optical flow constraint (OFC) equation has been extensively studied in computer vision to estimate motion from image sequences. Traditionally the OFC relies on spatio-temporal brightness variations caused by motion. However, many other features are in principle appropriate, including gradient directions of the image intensity. Given gradient directions are more tolerant to changes in lighting, they seem to be an adequate choice for computing optical flows. Their applicability in the OFC is, however, not straightforward as the gradient direction is independent of the gradient magnitude, and is not defined in homogeneous areas. To palliate to these difficulties, a form of OFC equation based on probability distributions of gradient directions is proposed. The performance of the approach is assessed by experiments realized on synthetic and real world images.