Use this URL to cite or link to this record in EThOS:
Title: Extending Minkowski norm illuminant estimation
Author: Troncoso Rey, Perla
Awarding Body: University of East Anglia
Current Institution: University of East Anglia
Date of Award: 2012
Availability of Full Text:
Access from EThOS:
Access from Institution:
The ability to obtain colour images invariant to changes of illumination is called colour constancy. An algorithm for colour constancy takes sensor responses - digital images - as input, estimates the ambient light and returns a corrected image in which the illuminant influence over the colours has been removed. In this thesis we investigate the step of illuminant estimation for colour constancy and aim to extend the state of the art in this field. We first revisit the Minkowski Family Norm framework for illuminant estimation. Because, of all the simple statistical approaches, it is the most general formulation and, crucially, delivers the best results. This thesis makes four technical contributions. First, we reformulate the Minkowski approach to provide better estimation when a constraint on illumination is employed. Second, we show how the method can (by orders of magnitude) be implemented to run much faster than previous algorithms. Third, we show how a simple edge based variant delivers improved estimation compared with the state of the art across many datasets. In contradistinction to the prior state of the art our definition of edges is fixed (a simple combination of first and second derivatives) i.e. we do not tune our algorithm to particular image datasets. This performance is further improved by incorporating a gamut constraint on surface colour -our 4th contribution. The thesis finishes by considering our approach in the context of a recent OSA competition run to benchmark computational algorithms operating on physiologically relevant cone based input data. Here we find that Constrained Minkowski Norms operi ii ating on spectrally sharpened cone sensors (linear combinations of the cones that behave more like camera sensors) supports competition leading illuminant estimation.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available