Date of Award

2005

Document Type

Master Thesis

Degree Name

Master of Engineering (Research)

Department

Department of Electronic Engineering

Abstract

Traditionally, the issue of colour constancy has limited the success of colour image segmentation algorithms. Numerous methods exist to achieve colour constancy for different applications. The principal contribution of this thesis is the investigation of a suitable technique for the problem of robust colour based object detection.

Appropriate colour models are investigated and analysed in order to assess their ability to enhance the isolation process. Models that possess suitable colour constancy characteristics are chosen for use with image segmentation algorithms. Reviews of existing sur\'eys of image segmentation algorithms are appraised for the emergence of a suitable colour image segmentation technique. Active contour models or snakes surface as being apposite. In particular, statistical snakes or, more fonnally, active region models have been demonstrated to possess desirable characteristics when applied to image segmentation. Specifically, statistical colour snakes have been shown to reduce the effects of colour constancy when a colour model, such as Hue Saturation Value (HSV) and normalised Red Green Blue (rgb) is adopted. It is the colour constancy characteristic of the colour model which dictates the statistical colour snake’s ability to isolate objects robustly under varying illumination conditions.

The algorithm’s performance therefore depends on the choice of colour space. The Tint Saturation Luminance (TSL) colour model has been reported as a suitable candidate for achieving colour constancy but has not been used in the context of a statistical colour snake. It is therefore used with statistical colour snakes on single and multiple images to evaluate its suitability as an inbuilt model for the snake. Comparative studies of the above mentioned colour models with statistical colour snakes is then carried out to determine the colour snake with the most robust behaviour in isolating an object of known colour under varying illumination.

Access Level

info:eu-repo/semantics/openAccess

Share

COinS