A Bayesian approach to object detection

Author

Nikulin, V.

Author_Institution

Div. of Marine Res., CSIRO, Hobart, Tas., Australia

Volume

2

fYear

1998

fDate

4-9 May 1998

Firstpage

809

Abstract

A binary hypothesis testing is a common tool in the task of object detection. Experiments with real images have confirmed the effectiveness of Neyman-Pearson detectors based on locally weighted sample mean and variance. In line with Bayesian approach the threshold parameter is defined as a function of prior distribution. According to the basic idea of Gibbs Sampler the neighbourhood system of image element determines the distribution of this element. Using the concepts described and taking any image as an initial we can form sequence of images in order to develop this dependence. As a result the quality of object detection can be improved significantly

Keywords

Bayes methods; computer vision; image classification; image sequences; object recognition; Bayes method; Gibbs Sampler; Neyman-Pearson detectors; binary hypothesis testing; image sequence; object detection; pattern classification; prior distribution; threshold parameter; Australia; Bayesian methods; Colored noise; Detectors; Gaussian noise; Object detection; Painting; Pixel; Signal to noise ratio; Testing;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks Proceedings, 1998. IEEE World Congress on Computational Intelligence. The 1998 IEEE International Joint Conference on

Conference_Location

Anchorage, AK

ISSN

1098-7576

Print_ISBN

0-7803-4859-1

Type

conf

DOI

10.1109/IJCNN.1998.685871

Filename

685871