Clothing colour clustering method based on CC_K-Means algorithm

Abstract

Colour matching is a very important part of clothing matching, which refers to the selection of single or multiple colours that match a colour for matching the whole or part of clothes. The key challenge is how to simplify it in the face of many colour combinations, otherwise a wide range of clothing colour combinations will make it more difficult to match the clothing colours. A CC_K-Means algorithm based on K-Means is proposed specifically for clothing colour clustering. The algorithm can be combined with a clustering learning curve to select the K-value with the better clustering effect from the unlabelled clothing colour types, so that the colours of the same cluster can be used for clothing colour matching using the centroid colour within the cluster, and the number of colour combinations is simplified to K*K A comparison test with K-Means, bisecting K-Means is conducted on a self-built 450 RGB clothing colour dataset. The results show that the method has a higher accuracy of clothing colour clustering, with an accuracy of 83.07%. This method not only ensures the accuracy of clothing colour matching but also reduces the number of colour pairs, which helps to simplify the process of clothing colour matching and contributes to other areas of colour matching.