Marr: vision as computation on three levels
In 1982 David Marr's book Vision appeared posthumously: vision in it is a sequence of representations running from a description of the image to a description of the three-dimensional objects around, and each problem has to be treated on three levels, the computational, the algorithmic and that of hardware implementation.
Why it matters
The division into levels of analysis gave neuroscience, psychology and computer vision a common language: ask separately what is computed and why, by what algorithm, and on what. By its publisher's account the book shaped a generation of brain and cognitive scientists.
Marr died in 1980; the catalogue lists the author as Marr, David, 1945-1980. The first edition was published by W. H. Freeman in San Francisco; the MIT Press reissued it on 9 July 2010. The book's own text was not read: the scan is in a lending library. So the record does not name the intermediate representations (primal sketch, 2.5-D sketch), does not claim that the book sets out the zero-crossing theory of 1980, and gives no month of publication.