COLOUR-DEPENDENT LINEAR VECTOR IMAGE FILTERING (WedAmOR3)

Author(s) : Stephen Sangwine (University of Essex, UK) Todd Ell (University of Essex, USA) Barnabas Gatsheni (None, UK)

Abstract : In previous work the authors have presented initial results from a new method of linear vector filtering applied to colour images. The method depends on resolution of colour pixel vectors into directions parallel and perpendicular to a chosen colour, followed by application of separate filters to the resolved image components in order to filter the chosen colour only (for example with a low-pass filter). In this paper we review this scheme and then present a new and detailed analysis of the limitations of the scheme as currently defined. Understanding of these limitations reveals for the first time where progress must be made with new linear operations before linear vector filters can be fully realized for practical use.

Menu