Browse Bar: Browse by Author | Browse by Category | Browse by Citation | Advanced Search
Author(s): Grossberg, S. | Wyse, L.L. |
Year: 1991
Citation: Neural Networks, 4, 723-742
Abstract: A neural network model, called an FBF network, is proposed for automatic parallel separation of multiple image figures from each other and their backgrounds in noisy gray-scale or multi-colored images. The figures can then be processed in parallel by an array of self-organizing Adaptive Resonance Theory neural networks for automatic target recognition. An FBF network can eparate the disconnected but interleaved spirals that Minsky and Papert introduced in their book Perceptrons. The network?s design clarifies why humans cannot rapidly separate interleaved pirals, yet can rapidly detect conjunctions of disparity and color, or of disparity and motion, that distinguish target figures from surrounding distractors. Figure-ground separation is accomplished by iterating operations of a Feature Contour System (FCS and a Boundary Contour System (BCS)-E?CS operations include shunting nets to compensate for variable llumination and diffusion nets to control filling-in. The BCS operations include oriented filters joined to competitive and cooperative interactions designed to detect, regularize, and complete boundaries in up to 50 percent noise, while suppressing the noise. hence the term FBF-that have been derive d from an analysis of biological vision....
Topics:
Biological Vision,
Image Analysis,
Machine Learning,
Models:
Boundary Contour System,
Author(s): Carpenter, G.A. | Grossberg, S. | Mehanian, C |
Year: 1989
Citation: Neural Networks, 2, 169-181.
Abstract: A neural network architecture is outlined that self-organizes invariant pattern recognition codes of noisy images. The processing stages are figure-ground separation, boundary segmentation, invariant filtering, and self-organization of a pattern recognition code by an ART 2 network. The article describes a new circuit for boundary segmentation, called the CORT-X filter, that detects, regularizes, and completes sharp (even one-pixel wide) image boundaries in up to 50% noise, while simultaneously suppressing the noise. The CORT-X filter achieves this competence by using nonlinear interactions between multiple spatial scales to resolve a design trade-off that exists between the properties of boundary localization, boundary completion, and noise suppression. The processing levels of the COR T-X filter are analogous to those of the Grossberg-Mingolla Boundary Contour System, but contain only feedforward operations that are easier to implement in hardware. The network nodes in these levels are analogous to cortical simple cells, complex cells, hypercomplex cells, and unoriented and oriented cooperative cells.
Topics:
Image Analysis,
Machine Learning,
Models:
ART 2 / Fuzzy ART,
Boundary Contour System,