A computer model of the initial stages of mammalian pattern processingShow full item record
Title | A computer model of the initial stages of mammalian pattern processing |
---|---|
Author | Zinser, Otto |
Date | 1970 |
Genre | Dissertation |
Degree | Doctor of Philosophy |
Abstract | The purpose of this investigation was the development and assessment of a computer model of the initial stages of mammalian pattern processing. The major basis for the model's development was Hubel and Wiesel's (1963) data, and their conception of the mammalian visual system. At the retinal and geniculate levels, receptor field systems were shown to respond to spots, at the cortical level, receptor field systems were described as being sensitive to lines or edges of particular orientations. Hubel and Wiesel speculated that the mammalian visual system is a hierarchical processor, such that sets of ganglion cells supply information to geniculate cells and sets of geniculate cells feed information to cortical cells. Hubel and Wiesel's conception of the functioning of the initial phases of the mammalian visual system was programmed for a computer. An analog of ganglion and of geniculate cell receptor fields was designed. Analogs of cortical edge-sensitive receptor fields were also prepared. Hierarchical processing was modeled by applying edge detecting analogs to the output of the spot analogs. Assessment consisted of having the model process nondegraded and degraded patterns. One form of degradation introduced random "noise," such as snow on a television screen: the other introduced blur, which concentrated on the destruction of the contour. A pattern assessment methodology was developed, which consisted of having Ss rate nondegraded, degraded, spot-analog processed, and combined, spot-analog processed and edge-analog processed versions by comparing them with the original. The results indicated that the spot-analogs degraded the nondegraded patterns to some extent but were quite effective in detecting the patterns from random "noise." They were less effective in detecting the blurred versions. The edge detecting analogs did not improve the appearance of patterns with either form of degradation; the edge detectors merely summarized the contour of the spot-detected version. It was concluded that the spot-analogs are most suitably applied to degraded patterns (random "noise," in particular), and the edge-detecting analogs are most suitably applied to nondegraded patterns. The results of this investigation suggested the possibility that spot detecting mechanisms in the mammalian visual system serve as a preprocessing system which detects patterns or which upgrades the quality of the image, whereas the edge detecting mechanisms may be part of a higher-level feature analytic process. It was suggested that the present model could be adapted to serve as a preprocessing phase of a general model of pattern recognition. |
Link | https://repository.tcu.edu/handle/116099117/34659 |
Department | Psychology |
Advisor | Evans, Selby H. |
This item appears in the following Collection(s)
- Doctoral Dissertations [1484]
© TCU Library 2015 | Contact Special Collections |
HTML Sitemap