Illumination compensation in textile colour constancy, based on an improved least‐squares support vector regression and an improved GM(1,1) model of grey theory

Abstract

In many colour‐related fields, such as textile colour measurement systems based on computer vision as well as dyestuff synthesis, colour constancy with illumination variation is a key problem to be solved. A new algorithm of illumination compensation for colour constancy for textured textiles is proposed in this paper, combining an improved least‐squares support vector regression (LSSVR) and an improved GM(1,1) model of grey theory. The LSSVR algorithm was improved by determining the error limitation according to the distance between a standard sample and a test sample, while the improved GM(1,1) model used the average value of the sequence as the initial iteration value. The disadvantage of LSSVR, that it easily falls into global optimisation, can be compensated for by the local optimisation capacity of the GM(1,1) model. Experimental results show that the algorithm of this paper is very stable and provides good illumination compensation, whereby the time complexity of compensation is reduced by repeatedly processing fractional data with the LSSVR algorithm, and the prediction accuracy is improved by combining the improved GM(1,1) model and the LSSVR algorithm.