Instantaneous Optical Flow

Abstract

So far in this book, we have been concerned with the development of a technique for segmenting images. However, as we have seen, this technique yields not only the segmented components of the constituent objects or images but it also extracts their global spatial velocities or disparities. In all of this, we have assumed that the scene comprises two distinct objects or images, each having a different velocity or disparity. However, if the scene which is being viewed doesn’t exhibit this figure-ground composition, we can still perform some useful work and establish the structure of the scene. How? Well, if we are able to compute global image velocity, then we should also be able to compute local image velocity, i.e., we should be able to use the technique to compute the local optical flow of the image. This in turn, together with some assumptions about the motion of either the observer or the objects in the scene, allows one to make inferences about the structure of the scene (see [159] for an excellent introduction to this subject). In this chapter we will show how the optical flow can be computed using a fairly well-established phase-based Fourier method. In the next chapter, we look at the possibility of deploying the full theory developed in this book and we’ll consider the circumstances under which it would be appropriate to do so.