Application of k -Means Clustering to Material Research: Measurement of Layer Thickness and Contact Angle

A new method of measuring material properties such as thickness or contact angle from recorded images is presented. A usual inspection of the image largely depends on the human decision for the boundary determination, which could induce the result to be biased. The present study proposes to use “ k -means clustering”, which is a method of vector quantization. The image was digitized, and the digitized data were categorized into several clusters (three in case of the thickness measurement, two in case of the contact angle measurement) as objective as possible. It was shown that this method is reliable and does not require any human decision, except for the number of clusters. The low-resolution image was well treated by the present method, which was not possible in other conventional methods. Graphic Abstract