Real World Image Analysis

Abstract

This thesis considers some of the image processing problems in trying to construct an automated system for detecting near-shore water-borne pollution (oil slicks) using land mounted visible band cameras. In particular, we develop a novel approach to the uniform scene illumination problem to retrieve reflectance more accurately, prior to segmentation. We also introduce a simple Kalman filter approach to exploit some of the dynamic information content across images through time to improve the slick segmentation through a probabilistic model.

Date of Award	2003
Original language	English
Awarding Institution	Aston University