No abstract
The past 15 years have been a productive, if not terribly illuminating, period for visual physiologists. Spurred on by the early electrophysiological successes of Hubel and Wiesel, we in the field had the feeling that the understanding of human vision was closer at hand. Early feline data were gradually replaced by primate data as the monkey became the model of choice. During this period, important advances were made in the neural bases of color, form, and binocular vision. It was surely, we thought, only a minor step to find the right connections between appropriately oriented simple, complex, and hypercomplex cells and the monkey hand neurons. It seemed as though virtually every known parameter relevant to the development of cortical binocularity had been thoroughly tested.At the same time, sine-wave fever was sweeping physiology and psychophysics, forcing researchers to trade in their Carousel projectors for oscilloscopic displays. There were also major breakthroughs in the application of axonal tracing techniques to the visual pathways. Many of the "established" facts in neuroanatomy have been since modified. Never before has our knowledge been better of gross internuclear connections, as well as the microorganization of the projection neurons.Into this complacent scientific milieu comes David Marr's book Vision, destined, I believe, to become one of the most revolutionary books of the decade. While Marr's other contributions (e.g., A theory for cerebral neocortex, 1970) are as original as Vision, they are highly technical and difficult to test empirically. In contrast, Vtsion is a formal, eminently testable, but nontechnical computational theory of human vision. Only one of its major contributions is its novel synthesis of the concepts of neural feature detection and sine-wave analysis.The major goal of computational vision is the design of a machine that will see. This goal forces Marr and colleagues to logically dissect visual behavior into distinct and novel levels of analysis.He warns us not to identify the computational levels with the typically studied synaptic levels, as cooperative neural networks are probably more important for global computations. Thus free of anatomical restraints, the theory appears to be highly successful. It is fertile ground for psychophysicists interested in testing various theoretical predictions, but may leave the physiologist looking for neural confirmation somewhat mystified.The author approaches the study of visual behavior with three major questions: (1) What is the goal of the computational problem and why is it appropriate? (2) How can the input and output of each stage of analysis be represented, and what is the algorithm for the transformation? (3) How can the representation and algorithm be realized physically? Little space is devoted to the hardware problems of question 3. For this reviewer, the most interesting question was the second. Marr divides the visual process into four stages. The most primitive stage is the two-dimensional retinal light intensity distri...
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.