Unsupervised Bayesian method for colour matching of product packaging

Abstract

There are some problems in the traditional colour matching methods of product packaging, such as high accuracy of colour matching and long consumption time. This paper proposes a colour matching method of product packaging based on unsupervised Bayes. First, the distance threshold between different colours of product packaging was obtained. Second, the colour features were extracted by sparse representation method, and the colour features were preprocessed by fuzzy c-means clustering. Third, the colour of product outer packaging is sorted according to the proportion to realise the quantification of product outer packaging colour. Then, the colour matching parameters of product packaging are determined by Dirichlet distribution in unsupervised Bayesian. Finally, the optimal colour matching scheme is determined by introducing multiple beta functions. The importance weight of matching scheme is modified by approximate posterior probability to complete the matching. The results show that the minimum error of the proposed method is about 0.5%.