Attention-focusing artificial neural networks for image analysis

Barker, S.E., 2000. Attention-focusing artificial neural networks for image analysis. PhD, Nottingham Trent University.

Preview

Text 10290309.pdf - Published version Download (36MB) | Preview

Abstract

Some of the common operations humans take for granted, for example the human vision system, have been found very difficult to emulate. Although humans are able to perceive visual information almost instantly, this belies the complexity of this process.

This thesis describes a computer vision strategy that involves Artificial Neural Networks (ANNs) to perform accurate and efficient object identification. Face location is used as the primary test domain. This involves the processing of real world scenes to distinguish between faces of different shapes, sizes and different viewpoints. Object identification in a complex environment is an extremely difficult task and research into this area of computer vision is currently not being fully exploited. Many previous models for computer vision have applied techniques that only solve particular well-defined problems.

An efficient two-stage vision strategy is presented which removes the necessity to process an image at full resolution through the use of low resolution. The first stage uses a multi-resolution approach to identify areas of interest at an optimally low resolution. The focus areas are then passed to a classification stage to perform more accurate analysis to reject the area of interest or confirm the presence of the pre-determined object.

Item Type:	Thesis
Creators:	Barker, S.E.
Date:	2000
ISBN:	9781369325584
Identifiers:	Number Type PQ10290309 Other
Divisions:	Schools > School of Science and Technology
Record created by:	Laura Ward
Date Added:	06 Jul 2021 09:29
Last Modified:	10 Apr 2024 14:56
URI:	https://irep.ntu.ac.uk/id/eprint/43335

Actions (login required)

Edit View

Views

Downloads