Texture Discrimination and Binding by a Modified Energy Model
Abstract
The visual system has a remarkable ability to discriminate subtle differences in texture patterns. Psychophysical studies have shown that texture discrimination occurs preattentively: namely, it operates in parallel over large regions of the visual field, occurs early in visual processing, and is unable to make distinctions based on multiple conjunctions of features. Recent approaches to understanding texture discrimination have followed two pioneering models. Julesz's texton model proposed that the visual system detects a relatively small number of primitive texture elements, called textons. Textons are features such as size, color, orientation, line endings, and junctions which, for the most part, are also the primitives to which visual cortical cells are selective. Julesz has shown that textures which differ in the density of one or more textons are distinguishable by human observers. In fact, textures which have identical second order statistic (i.e., identical Fourier transforms) and even identical third order statistics are still distinguishable if they differ in texton density. A second recent approach to texture discrimination is the use of energy models. The basic idea of these models is to sample, at several spatial frequencies, the amount of stimulus energy present (energy is loosely defined as the averaged squared output of a set of detecting elements). Such models are well-suited to network implementation and have been shown to work well in a variety of cases.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1992
- Accession Number
- ADA248130
Entities
People
- Ko Sakai
- Leif H. Finkel
- Paul Sajda
Organizations
- University of Pennsylvania