Optical flow cannot be computed locally, since only one independent measurement is available from the image sequence at a point, while the flow velocity has two components. A second constraint is needed. A method for finding the optical flow pattern is presented which assumes that the apparent velocity of the brightness pattern varies smoothly almost everywhere in the image. An iterative implementation is shown which successfully computes the optical flow for a number of synthetic image sequences. The algorithm is robust in that it can handle image sequences that are quantized rather coarsely in space and time. It is also insensitive to quantization of brightness levels and additive noise. Examples are included where the assumption of smoothness is violated at singular points or along lines in the image.

We present a new approach to computing from image sequences the two-dimensional velocities of moving objects that are occluded and transparent. The new motion model does not attempt to provide an accurate representation of the velocity flow field at fine resolutions but coarsely segments an image into regions of coherent motion, provides an estimate of velocity in each region, and actively selects the most reliable estimates. The model uses motion-energy filters in the first stage of processing and computes, in parallel, two different sets of retinotopically organized spatial arrays of unit responses: one set of units estimates the local velocity, and the second set selects from these local estimates those that support global velocities. Only the subset of local-velocity measurements that are the most reliable is included in estimation of the velocity of objects. The model is in agreement with many of the constraints imposed by the physiological response properties of cells in primate visual cortex, and its performance is similar to that of primates on motion transparency. 1.

We show how to actively compute qualitative shape properties directly from image intensities - that is, without having to first "reconstruct the surface". Our approach diverges from classical active shape from shading [1] in two important ways. First, we do not attempt to compute a dense depth map of the surface. We rather detect the presence of certain qualitative geometric features. Second, we use two different types of lighting conditions rather than two different instances of a single type (the point source). Our two types of lighting conditions are a diffuse source and a point source positioned at the camera. The major result of this paper is that concave orientation discontinuities and smooth valleys share the same shading signatures. Under diffuse conditions, both produce local minima in the image intensity. Under the point source condition, both produce local maxima in the image intensity. One can detect these features by alternating between the lighting conditions. 1 Introdu...

How may the shape of a smooth surface be inferred from an image? Traditional methods in computer vision for inferring shape-from-shading assume that surface shading depends entirely on surface orientation. In many illumination scenarios, however, shading may occur independently of surface orientation. For example, when an extended light source such as the sky casts a continuous shadow on a flat ground, the resulting shading is due to illumination variations only. In this thesis, a new approach to shape-from-shading is taken in which shading variations are attributed entirely to spatially varying illumination. This leads to a new analysis of the shape-from-shading problem, and to a new algorithm for solving it. In particular, a model of spatially varying illumination is developed which is in terms of the set of light rays in free space. This set is shown to be a four dimensional smooth manifold, called the radiation manifold. Local transformations between coordinate systems on this mani...