A random-field model-based algorithm for anomalous complex image pixel detection

Author

Bello, Martin G.

Author_Institution

Charles Stark Draper Lab., Cambridge, MA, USA

Volume

1

Issue

2

fYear

1992

fDate

4/1/1992 12:00:00 AM

Firstpage

186

Lastpage

196

Abstract

Random-field model-based algorithms for the detection of anomalous pixels associated with complex valued imagery may be essential to robust focus of attention, target detection, and curing. The described algorithm includes the fitting of a specific class of causal, two-dimensional autoregressive random-field models to image data over specified estimation windows, and then subsequent construction of prediction error samples over specified detection windows. Statistical testing of the calculated prediction error samples is then used to localize anomalous image pixels. Experimental results obtained from running the described algorithm on SAR (synthetic aperture radar) imagery are included

Keywords

picture processing; signal detection; 2D AR models; SAR imagery; anomalous complex image pixel detection; causal models; curing; detection windows; estimation windows; image data; prediction error samples; random field model algorithm; synthetic aperture radar; target detection; Computer vision; Cultural differences; Focusing; Helium; Object detection; Object recognition; Pixel; Predictive models; Robustness; Statistical analysis;

fLanguage

English

Journal_Title

Image Processing, IEEE Transactions on

Publisher

ieee

ISSN

1057-7149

Type

jour

DOI

10.1109/83.136595

Filename

136595