2 November 2011 A novel illumination-invariant colour constancy algorithm
Author Affiliations +
The human brain is able to extract the colour of objects no matter the existing illumination conditions which may affect the appearance of their colour. For a person, an object that is red, it will be red regardless of the type of illumination source. However, for a computer algorithm to achieve the same task is not as simple. For vision and robotics applications were feature extraction is essential having such an algorithm is crucial. It is well known from the literature that the colour of an object does not only depend on the chemical composition and shape of its surface but also on the illumination conditions, the intensity, number, location and colour of the sources of illumination as well as the intrinsic and extrinsic characteristics of the sensor used. Based on that knowledge, we propose a simple colour constancy algorithm that uses the quaternion representation of each pixel in the image instead of the commonly used RGB model. We assume linearity in the reception of the luminance spectrum of the charge-coupled device of the camera before variations in the illumination. We have tested our colour constancy algorithm in a variety of images containing different colour objects under different illumination conditions. Our experimental results show the feasibility of the proposed method.
© (2011) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
L. A. Torres-Méndez, L. A. Torres-Méndez, M. L. Quiñones Muñoz, M. L. Quiñones Muñoz, E. J. Olaya-Benítez, E. J. Olaya-Benítez, } "A novel illumination-invariant colour constancy algorithm", Proc. SPIE 8011, 22nd Congress of the International Commission for Optics: Light for the Development of the World, 80119C (2 November 2011); doi: 10.1117/12.901951; https://doi.org/10.1117/12.901951


A watermarking algorithm for map and chart images
Proceedings of SPIE (March 21 2005)
Improved boundary tracking by off-boundary detection
Proceedings of SPIE (November 08 2012)
Multiresolution Object Detection And Segmentation
Proceedings of SPIE (September 26 1989)

Back to Top