|
Accession number;99A0938065
|
| Title;A Pattern recognition using Maximal relation principle. |
| Author;
SUZUKI AKIO
(Tohoku Univ.)
|
Journal Title;Record of Electrical and Communication Engineering Conversazione, Tohoku University
|
Journal Code:F0511A
|
ISSN:0385-7719
|
|
VOL.68;NO.1;PAGE.2-4(1999)
|
| Figure&Table&Reference;FIG.3, REF.7 |
| Pub. Country;Japan |
| Language;Japanese |
| Abstract;In this paper, I propose a model that can separate arbitrary figures which include illusory contours from visual scene. Figure must be made to relation among features because it is impossible to distinguish signals from noise. To extract boundaries for segregating figures, symmetric and antisymmetric filters are required. Each filters constraint the other and can create structure of input. Furthermore filters with various spatial frequency tunings are required as features to segregating spatial pattern. These filters are supported by psychophysics and physiological experience. In this model, these filters process input signals in parallel and arbitrary figures are flexibly and uniquely formed by integration outputs from filters using miximal relation principle. (author abst.) |
|
|
|
Related Articles;
|
