MIT Progress in Understanding Images

Abstract : In this series of Image Understanding Workshop Proceedings, we have stressed the issue of representation. In particular, we have described the development by Horn and his colleagues of the reflectance map, the albedo image, and the Gaussian image, and we have described the work of the group founded by Marr using the primal sketch, the 2 1/2 D sketch, and axis based 3-D models. In the April, 1981 Proceedings, we reviewed work on computing shape from shading and occluding boundaries, including a local parallel algorithm for doing this due to Ikeuchi and Horn 1981; the detection and perception of motion by computing the optical flow and directional selectivity of zero crossings; the interpolation of curves and surfaces; the real-time convolution of images with a difference-of-Gaussians (DOG) operator; and progress toward computing the full primal sketch. Here we review work on stereo to facilitate the computation of depth information and visible surface characteristics, the detction and interpretation of motion, the interpolation and description of visible surfaces, the description of two- and three-dimensional shapes, real-time convolution, and shape from shading. (Author)