We propose and evaluate a class of objective functions that rank hypotheses for feature labels. Our approach takes into account the representation cost and quality of the shapes themselves, and balances the geometric requirements against the photometric evidence.
A single two-dimensional image is an ambiguous representation of the three-dimensional world–many different scenes could have produced the same image–yet the human visual system is extremely successful at recovering a qualitatively correct depth model from this type of representation.
Attempts have been made to speed up image-understanding computation involving conventional serial algorithms by decomposing these algorithms into portions that can be computed in parallel.
Nils J. Nilsson, J.M. Agin, Barbara G. Deutsch, Richard E. Fikes, Earl D. Sacerdoti, & J.M. Tenenbaum
This report describes the goals and plans for a five-year project to develop a computer-based system that will serve as an expert consultant to a human apprentice.
Earl D. Sacerdoti, Richard E. Fikes, Rene Reboh, Daniel Sagalowicz, Richard J. Waldinger, & B.M. Wilber
This paper presents a functional overview of the features and capabilities of QLISP, one of the newest of the current generation of very high level languages developed for use in artificial intelligence (AI) research.
In 1981 SRI introduced RANSAC, a now widely referenced paradigm for robust communication ideally suited to computer vision (a type of artificial intelligence used in image analysis.
In this paper, we present a method for registering images of complex 3-D surfaces that does not require explicit correspondences between features across the images.
This report summaries the results of a three-year project aimed at the design and implementation of computer languages to aid in expressing problem solving procedures in several areas of artificial intelligence including automatic programming, theorem proving, and robot planning.
