PubMed Health⌕ Search
Home/Archive/PubMed

PubMed · 8817773

Machine learning paradigms for pattern recognition and image understanding.

T Caelli·W F Bischof
View original source ↗Export metadata ↓

Abstract

In this paper some issues are considered related to the encoding of spatial information and associated perceptual learning algorithms which, it is claimed, are necessary for robust pattern and object recognition in multi-object (natural) scenes. The types of learning requirements within a 'recognition-by-parts' paradigm are contrasted with findings from alternative models.

Explore related subjects

Form PerceptionHumansLearningNeural Networks, ComputerPattern Recognition, VisualSpace PerceptionJournal ArticleResearch Support, Non-U.S. Gov'tReview

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

T Caelli, W F Bischof. 1996. Machine learning paradigms for pattern recognition and image understanding.. https://doi.org/10.1163/156856896x00079

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related citations

PubMed1998

Developments in generalized ranked-order filters.

A general formulation of ranked-order filters is developed in two parts: part 1, signal-to-noise-ratio analysis and part 2, construction and analysis of a ranked-order-filter function based on a mathematical logic approach. The filter function is analyzed to define the structure of filter roots for one-dimensional (1-D) and two-dimensional (2-D) window filters as data patterns that are invariant of the filter. The 1-D and 2-D coded window filters defined for roots of repeated patterns of binary data are defined and analyzed. The analysis concludes with an application of the coded window filter to a computer-generated 2-D noisy image containing a binary pattern and an application for feature extraction by a 2-D filter constrained by a predicate function to select only fixed-point root data structures.

Form Perception↗
PubMed1998

Bounds on null functions of linear digital imaging systems.

Any linear digital imaging system produces a finite amount of data from a continuous object. This means that there are always null functions, so a reconstruction of the object, even without noise in the system, will differ from the actual object. With positivity constraints, the size of a null function is limited, provided that size is measured by the integral of the absolute value of the null function. When smoothing is used in reconstruction, then smoothed null functions become relevant. There are bounds on various measures of the size of smoothed null functions, and these bounds can be quite small. Smoothing will decrease the effects of null functions in object reconstructions, and this effect is greater if the smoothing operator is well matched to the system operator.

Form Perception↗
PubMed1997

Absence of contour linking in peripheral vision.

Human foveal vision is subserved initially by groups of spatial, temporal and orientational 'filters', the outputs of which are combined to define perceptual objects. Although a great deal is known about the filtering properties of individual cortical cells, relatively little is known about the nature of this 'linking' process. One recent approach has shown that the process can be thought of in terms of an association field whose strength is determined conjointly by the orientation and distance of the object. Here we describe a fundamental difference in this feature-linking process in central and peripheral parts of the visual field, which provides insight into the ways that foveal and peripheral visual perception differ. In the fovea, performance can be explained only by intercellular linking operations whereas in the periphery intracellular filtering will suffice. This difference represents a substantial economy in cortical neuronal processing of peripheral visual information and may allow a recent theory of intercellular binding to be tested.

Form Perception↗